The Red Giant, by Rufus Curnow

Why you can't be AI literate without being data literate

Thu, 30 Jul 2026 09:00:00 +0000

Data literacy is a frequently overlooked part of AI literacy. If your AI literacy training doesn’t cover data then that’s a common gap. In DataCamp’s 2026 State of Data and AI Literacy report, a survey of more than 500 US and UK leaders, 88% said basic data literacy matters for their people’s day-to-day work, yet only 42% of their organisations provide training in it across the workforce. It’s worth understanding why data literacy matters so much to AI literacy.

The main reason is very simple: AI runs on data. The quality of anything an AI system produces depends on the quality of the data it was built on and the quality of the data you give it. If people can’t think clearly about that data, they won’t be able to think clearly about the AI either. Data literacy enables someone to use AI tools well and to judge the outputs.

It helps to clarify what data literacy actually is, because it’s often misunderstood as being good with numbers or comfortable in a spreadsheet. It’s broader than that, because data includes the emails in your inbox, the notes from a meeting, a set of customer reviews, the photos in a shared drive, and the contracts in a folder somewhere. Being data literate means being able to make sense of all of it: where it came from, what it actually tells you, what it leaves out, and what you can trust.

For an organisation, data literacy typically shows up as groundwork for AI literacy. AI is only as useful as the data it can reach, so a data-literate organisation:

tidies up and clears out duplicate files that could provide AI tools with incorrect or outdated data
makes sure AI tools can get to the data they need
sets permissions carefully, so employees can’t get to the data they shouldn’t

This data literacy groundwork isn’t glamorous, but it’s a critical part of enabling organisational AI capability.

There’s a competitive edge in the data groundwork too. Every organisation can buy the same AI tools, so the tools themselves give you nothing your competitors can’t have. Your data is the one thing competitors don’t have, which makes it worth treating as a valuable asset.

For an individual using AI day to day, data literacy is more about judgement. When someone uses an AI tool well, they’re questioning the data twice over. On the way in they question what information to give the AI, because a model can only reason about what it can see. And on the way out they question what comes back. If you ask an AI to summarise customer feedback but only feed it last month’s reviews, it will give you a confident summary of a partial picture, and unless you know what it was working from, you won’t spot the problem.

Part of that judgement is knowing what you should and shouldn’t put into a tool in the first place. Feeding confidential or personal data into a public AI tool can mean handing it to a third party, with all the security and confidentiality risks that carries, so recognising that is itself a data literacy skill.

Bias is a data problem too, because an AI system trained on skewed data will produce skewed results, often without anyone noticing. I’ve written a separate article about bias in AI, so I won’t repeat it here, except to say that spotting it depends on the same data literacy skill, the ability to ask what the underlying data represents and who or what it leaves out.

Data literacy also shapes a decision leaders now face regularly: whether a given use of AI is worth paying for. AI is increasingly priced by usage, so working out whether AI use is justified means reviewing the data on usage, time saved and value returned. Without that literacy, it’s easy to keep paying for AI that isn’t providing a return on investment, or to drop AI where it’s adding value.

So when you’re scoping AI literacy training, look at whether it covers data or jumps straight to the tools. What does covering data actually look like? It should be practical rather than technical: where the organisation’s data lives and what state it’s in, what should and shouldn’t go into an AI tool, how to judge what comes back, and how to handle confidential and personal information along the way.

The organisations getting the most from AI treat data literacy and AI literacy as connected capabilities, built at every level, because in practice that’s what they are. The evidence points the same way: in the same DataCamp report, 21% of organisations reported significant returns on their AI investment, but among those with a mature data and AI literacy programme across the whole workforce, that doubled to 42%.

Which AI tools and models should you use?

Fri, 12 Jun 2026 09:00:00 +0000

The question I probably get asked more than almost any other is some version of "what's the best AI tool for me to use?" People typically want a single answer and I understand why, but the honest answer is that there isn't a single tool or model that’s best for all use cases (at least at the time of writing). Different tools and different models have different strengths and weaknesses, and learning about their capabilities is a key part of AI literacy. That's true across the big tools (ChatGPT, Claude, Copilot and Gemini), and it's also true within them, because most tools give you a choice of models.

When you get familiar with the capabilities, the simple rule of thumb I suggest is to match the tool to the job, and if you're not sure which level model to use, start with a lower one and move up only if it doesn’t give you what you need.

Match the tool to the job

Over time you'll build a feel for which tool suits which kind of work, and you'll get there either by trial and error or advice from others. I've landed on my own preferences through a lot of trial and error with the tools, but I can afford to prioritise testing the tools because it’s a key requirement of my job. Currently I use Claude when I'm writing or thinking something through, ChatGPT or Gemini when I'm doing research or generating images, and Copilot is great for looking across the Microsoft ecosystem. Those are my preferences today, but new models are coming out more frequently than ever, and the capability of the tools is regularly increasing. For most people it's not practical to keep switching to the new 'best' tool, and it only makes sense to switch if there's a significant shift in capability. What you should do is explore the capabilities of the tool you have access to, and you’ll discover where there are use cases requiring access to another tool. The skill is understanding which tools work best for you in which circumstances, and the value comes from learning how to use a tool which is at least 'good enough' well rather than consistently switching to the latest ‘best’ tool.

Start light, and only move up if you need to

The second half of my rule of thumb is about which model to use within a tool. Most tools now offer a range from quick response models through to slower, more capable thinking models. I suggest avoiding the temptation to always select the most powerful option because it’s often unnecessary and inefficient to do so.

In practice the faster, lighter models are good enough for a surprising amount of everyday work, and they give you the answer in a fraction of the time. My advice is to test tasks on a lighter model first, and if it does the job to a sufficiently high quality, you have no need to move up. If the output falls short, that's your signal to try a more powerful model. Do this for a while and you'll quickly learn which kinds of task really need the more powerful models and which don't, and that judgement is becoming an increasingly valuable skill.

There's a practical reason to get this right that goes beyond speed. The more capable models don't just take longer, they also eat into your usage more quickly, and most tools are now either putting limits on how much you can use in a given window, or in some cases they're starting to charge on a usage basis. Using your most capable models on a job a lighter model could have handled is a waste of that allowance or money, so learning which model to use is about being efficient with your usage, and knowing that the greater capability is there when you truly need it.

The value in using more than one tool

When you’ve been using AI for a significant period of time, you’ll probably discover that you want access to more than one tool. It happens because different tools simply suit different jobs, and it can also be valuable to use more than one tool so that you can cross-check and compare responses. You wouldn’t want to do that for every task, but for some tasks it can be very valuable. For a piece of research, you can run the same question across several tools, say Gemini, ChatGPT, Claude and Perplexity, then collate the answers and look at where they agree and where they differ. The points they agree on are the ones you can be most confident in, and the places they diverge are where you should dig deeper and spend more time checking. Copilot even has this built in now with their Critique and Model Council options which run queries on both GPT and Claude, and using multiple models is one of the better defences I know against hallucinations.

The same logic scales up to organisations, and it's worth saying plainly: Not everyone needs the same set of tools. You should start with a tool that everybody knows how to use, but some people will need access to additional tools for different use cases.

The tools will keep changing faster than most people can keep up, and the specific references I’ve made here will date, but developing a habit of matching the tool to the job, starting on a lower capability model, and building your judgement to determine if you need a more capable model, is a skillset that will still serve you long after today's models have been replaced. That skill will separate people who use AI well from those who just use it.

What good AI training actually looks like

Sun, 24 May 2026 09:00:00 +0000

I believe everybody should receive at least basic AI literacy training. The market for delivering training is filling up fast and becoming more competitive, so how do you tell the good from the average, and what should good AI training actually look like?

I'll acknowledge the self-interest of this article up front: I train people on AI for a living, so this article is partly an argument for hiring me. I've delivered AI training to over 2,000 people across sectors including higher education, financial services and construction, which gives me a basis for the checklist below. However, this checklist should also serve as a list of things to look for whoever you get your AI training from.

It starts with understanding you

The single biggest predictor of whether training delivers value is whether you and the trainer have aligned on what it will cover. That means a proper conversation beforehand: what tools your people have access to, what the regulatory and commercial context is, what they already know, and what you want them to be able to do afterwards. The exercises, the examples, and the framing should be shaped around your world. Much of the underlying theory will be consistent, but people need to know that the training is relevant to their roles and their industry, and it's not too much to expect a trainer to have a conversation with you and do some research on your industry in advance.

It's live and interactive

For AI training, learning how to use AI from videos that quickly become outdated doesn't really get you very far. AI is something you learn best by doing, and that requires live training with someone who can keep people on track, adjust the pace as necessary, answer the questions they're trying to ask, and notice when people haven't been able to follow the exercises. In the workshops I run, I spend as much time as possible with people on their own laptops, trying things and giving them help when they get stuck. In person is ideal because you can walk around and look at people's screens, but I acknowledge that live online is often more practical to arrange.

It's current

AI is moving fast and the pace of change is accelerating. The tools, the capabilities, and the emerging consensus on what's working in businesses shift month by month and week to week. Good training reflects where things are at the time of delivery. That means working with someone who is actively using these tools themselves and refreshing their material constantly. A straightforward way to test this: ask a prospective trainer what they've changed in their material in the last month, or which tool they've been using most this week. If they struggle to answer, that tells you something.

It covers the risks and the opportunities

Good training doesn't just teach people how to use AI. It helps them understand what it means: for their sector, their organisation, and their role. That includes helping people uncover and understand opportunities where AI can meaningfully accelerate work, which use cases are proving out in their industry, and the competitive implications. Good training should also cover the risks associated with AI: shadow AI (staff using personal AI tools at work without oversight), data protection and compliance exposure, over-reliance, and the limitations of the tools. People should leave understanding what AI can't do just as clearly as what it can. That calibrated judgement is more valuable than a list of generic prompts.

It builds on your expertise

A lot of AI training focuses on valuable skills like how to prompt, but that doesn't help if the training doesn't also help you understand how to use AI well in your specific role, alongside the expertise you already have. The real value comes from learning to use AI in a way that builds on what you already know: interrogating the output rather than accepting it, applying your own expertise to shape and challenge what AI produces, and building the judgement to know where it helps and where it doesn't.

It's engaging and meets people where they are

Pretty much every training session I've run has included people with a wide range of experience and opinions about AI. In any training session there are typically people who are already enthusiastic about AI, and people who are nervous about it, people who regularly use AI, and people who rarely use it. Good training should cover all of these groups. It keeps the enthusiasts engaged without leaving the sceptics behind, and it takes concerns seriously rather than brushing them aside. It should also, frankly, be fun. People learn better when they're engaged and retain more when they enjoy themselves. The best sessions are the ones where people are still talking about what they did weeks later.

It's designed for what happens next

You can't learn AI in a day. The people who develop real capability are the ones who keep going after the session: practising, experimenting, applying what they covered to their actual work. Good training is designed with that in mind. People leave with very clear next steps, specific things to try grounded in their own role, and ideally a way to come back with questions as they practise. A session with no follow-through and no expectation of practice will leave most people roughly where it found them within a fortnight.

That's my starting checklist. There's plenty more that goes into a good session: how to design exercises that build progressively, how to ensure inclusivity, how to balance group work and individual practice, but if your training provider can give you honest, specific answers to the things above, you're in good shape. If they can't, keep looking.

Shadow AI, and why banning tools isn't the solution

Sun, 10 May 2026 09:00:00 +0000

What is Shadow AI?

Right now, employees at your organisation are almost certainly using AI tools you haven't approved. An employee may be pasting a client brief into ChatGPT. A colleague in finance may be using a free AI tool to draft a report. Someone in HR may be asking Claude to summarise a set of CVs. None of them were doing anything intentionally malicious or dangerous, but none of them asked for permission.

This is shadow AI: the use of AI tools that haven't been approved, vetted, or sanctioned by the organisation.

The scale of the problem

Multiple pieces of research show that this is a significant and growing problem, and the scale of it may surprise you. Microsoft's UK research, published in October 2025 and based on a survey of 2,003 UK employees, found that 71% have used unapproved consumer AI tools at work, and 51% do so every week.

This isn't primarily a story about employees who don't know what they're doing or who are deliberately flouting the rules. In most cases, people are simply trying to do their jobs better and more efficiently, using tools they know work. Research by BlackFog, based on a survey of 2,000 UK and US employees conducted by Sapio Research in November 2025, found that 60% of employees accept the security risks of unsanctioned tools because they help them work faster or meet deadlines. 28% of UK employees using unapproved tools told Microsoft the reason is simply that their organisation doesn't provide a sanctioned alternative. There's no malice here, just momentum and a lack of approved alternatives. Understanding these reasons matters if you want to respond to it effectively.

The organisational risks

The problem is that the stakes are considerable. When an employee pastes a client brief or a financial dataset into a consumer AI tool, that data leaves the organisation's control entirely. It's transmitted to a third-party provider, processed on infrastructure the organisation doesn't own, and in some cases retained for model training purposes. The Samsung incident in 2023, in which engineers fed proprietary source code into ChatGPT for debugging help, only to have that code potentially surface in responses to other users, is the most widely cited example of how quickly this can go wrong.

IBM's Cost of a Data Breach Report 2025 found that shadow AI contributed to 20% of all data breaches and added an average of $670,000 to breach costs. IBM's IBV report "Go further, faster with AI," published in December 2025, found that 62% of senior executives say shadow AI risk has increased over the past two years. The risks don't stop at data either, with regulatory exposure, IP loss, reputational damage, and compliance failures all on the table when AI tools operate outside any governance framework.

Why banning unapproved tools doesn't work

The instinctive organisational response to shadow AI is to prohibit it. Block the domains, issue a policy, send a reminder about acceptable use. It's understandable, but I’m sorry to say, it doesn't work.

The Awareways Trend Report 2025, drawing on over 38,000 responses from employee training sessions, found that 89% of employees know they're supposed to seek approval before using new software, yet fewer than half actually do so.

You can write the policy, and most of your workforce will nod and carry on as before, just with a little more discretion about it. The BlackFog research reinforces the point from a different angle: 49% of employees are already using unsanctioned AI tools, and that figure is drawn from organisations with more than 500 employees, where policies and IT oversight are more likely to exist.

Banning tools doesn't just fail to eliminate shadow AI, it drives it underground, which is the worst possible outcome, because at least visible use can be monitored and addressed. When the Microsoft research found that only 32% of employees using unapproved AI tools are concerned about data privacy, and only 29% about IT security, the lesson isn't that employees are careless. It's that they haven't been given the context, the tools, or the alternatives that would make responsible use the path of least resistance. The real cause of shadow AI isn't employee recklessness, it's the governance gap, and banning tools doesn't close it.

What good governance actually looks like

The practical response to shadow AI isn't prohibition, it's making the legitimate route easier than the illegitimate one. That means:

having a clear, published AI acceptable use policy that explains what data can and can't be shared with which tools.
giving employees access to enterprise-grade AI tools that actually meet their needs, because if the sanctioned option is good enough, most people will use it.
naming someone who owns AI governance, so there's a clear point of accountability and a person employees can go to with questions.
building a swift, accessible approval process for new tools, so that when a team finds something useful, there's a realistic route to making it official quickly.

None of this requires a lengthy compliance programme before anything moves. Start with a few high-risk categories, define clear rules around data handling for those, and build from there. IBM's IBV research found that organisations with strong governance report 52% improved time to value on AI projects, and that one in four unsuccessful AI projects stems directly from weak governance. The goal with good governance is a framework that's live and evolving, not a policy document that nobody reads.

The Red Giant says...

If you're a leader reading this, the question worth asking isn't "are my employees using unapproved AI tools?" The data shows they almost certainly are. The question is whether you've made it easy enough for them to use appropriate AI tools responsibly, or whether the path of least resistance in your organisation still runs straight through the governance gap.

Shadow AI exists because people are trying to do their jobs well, and AI helps them. The answer to that isn't restriction, it's direction: clear policies, good tools, named ownership, and a fast enough approval process that doing things properly doesn't feel like more trouble than it's worth.

References

Awareways (2026) Trend Report 2025: The Rise of the Invisible Colleague: Shadow AI. Author: Sjoerd van Veldhuizen, MSc. Co-authors: dr. Jan-Willem Bulée and Remy Dijkstra, MSc. Awareways B.V., Utrecht. Published March 2026. Available at: https://www.awareways.com

BlackFog (2026) Shadow AI Threat Grows Inside Enterprises. Research conducted by Sapio Research on behalf of BlackFog, Inc., November 2025. Published 27 January 2026. Available at: https://www.blackfog.com/blackfog-research-shadow-ai-threat-grows/ (Accessed: 8 May 2026).

IBM Institute for Business Value (2025) Go further, faster with AI: How governance increases velocity. IBM Institute for Business Value. December 2025. Available at: https://www.ibm.com/thought-leadership/institute-business-value/en-us/report/ai-governance-trends (Accessed: 8 May 2026)

IBM Security (2025) Cost of a Data Breach Report 2025. IBM. Available at: https://www.ibm.com/reports/data-breach (Accessed: 8 May 2026).

Microsoft (2025) Rise in 'Shadow AI' tools raising security concerns for UK organisations. Microsoft UK Stories. 13 October 2025. Available at: https://ukstories.microsoft.com/features/rise-in-shadow-ai-tools-raising-security-concerns-for-uk/ (Accessed: 8 May 2026).

Why every organisation needs to take AI safety seriously

Wed, 15 Apr 2026 09:00:00 +0000

Introduction: Why ignoring AI safety is no longer an option

Artificial intelligence is already being used inside organisations, whether leaders have approved it or not. In many cases, organisations have taken action by purchasing tools such as Microsoft Copilot or ChatGPT, rolling out licences, and offering basic training focused on productivity. What is often missing is equivalent attention to safety, governance, and how these tools influence decisions once they are embedded in everyday work.

The risks associated with this are no longer hypothetical. In Canada, a tribunal ruled that Air Canada was liable after its website chatbot provided incorrect information about bereavement fares, requiring the airline to compensate a customer (American Bar Association, 2024). Evidence also shows that AI use in the workplace is widespread. In late 2025, 45 per cent of U.S. employees reported using AI at least occasionally in their work (Gallup, 2025), and research indicates that sensitive corporate information is regularly shared with generative AI tools, often outside formal oversight (Harmonic Security, 2025).

The challenge for leaders is therefore not whether AI is being adopted, but how it is being governed. AI risk does not depend on whether an organisation has an AI strategy or approved tools. It depends on whether AI is used in ways that affect decisions, data, or actions. In some organisations, that threshold has already been crossed. In many others, it is at risk of being crossed quietly as AI becomes more embedded in everyday work.

In 2025, Swedish fintech company Klarna publicly acknowledged that its earlier strategy of replacing hundreds of customer service employees with generative AI agents had negatively affected service quality and customer satisfaction. After initially claiming that AI could do the work of many humans and reducing its workforce accordingly, Klarna’s leadership reversed course and began rehiring human staff to ensure customers always have a clear path to human support. The CEO admitted that the company had “gone too far” in prioritising efficiency at the expense of quality, prompting a strategic shift back toward hybrid human-AI service delivery (Economic Times, 2025; CX Dive, 2025).

What organisations can do

Organisations should treat speed as a risk factor in AI deployment decisions, not just a benefit. Before automating customer-facing processes or reducing human involvement in service delivery, set clear success criteria and agree in advance what would trigger a pause or reversal. Pilot at small scale before committing, and monitor outcomes that matter to customers and staff, not just operational metrics. Building in a deliberate review point before scaling is far less costly than unwinding a decision after problems have already become public.

Conclusion

AI safety failures rarely come from reckless leaders or deliberate misuse. They emerge from everyday use, growing reliance, and pressure to move quickly. AI tools become embedded in normal work before shared understanding and governance have caught up. The response is not to ban AI or assume that licences and basic training are enough. It is to make AI use visible, governed, and reviewable.

That starts with understanding where AI is already used and encouraging people to share how they rely on it in practice. It means setting clear ownership for outcomes, keeping policies simple enough to follow under pressure, and training people to recognise common risks rather than expecting deep technical expertise. The goal is not to slow adoption, but to prevent avoidable incidents that damage trust, reputation, or compliance.

Organisations that handle AI well share a clear pattern. They know where AI supports work and where human judgement must take precedence. They can explain how AI influences outcomes, who is responsible, and what happens when something goes wrong. This clarity reduces risk, but it also enables faster, more confident use of AI. AI safety, in this sense, is not a constraint on innovation. It is a core organisational capability. The question for leaders is whether to build it deliberately now, or reactively after an incident forces the issue.

References

American Bar Association (2024) ‘BC tribunal confirms companies remain liable for information provided by AI chatbot’, Business Law Today, 29 February. Available at: https://www.americanbar.org/groups/business_law/resources/business-law-today/2024-february/bc-tribunal-confirms-companies-remain-liable-information-provided-ai-chatbot/ (Accessed: 28 January 2026).

BBC News (2023) 'ChatGPT: US lawyer admits using AI for case research', BBC News, 27 May. Available at: https://www.bbc.co.uk/news/world-us-canada-65735769 (Accessed: 28 January 2026)

Gallup (2025) AI use at work rises, 15 December. Available at: https://www.gallup.com/workplace/699689/ai-use-at-work-rises.aspx (Accessed: 28 January 2026).

Harmonic Security (2025) ‘Workers are spilling secrets to chatbots’, Axios, 31 July. Available at: https://www.axios.com/2025/07/31/workers-company-secrets-chatgpt (Accessed: 28 January 2026).

Responsible AI Labs (2024) AI hiring bias legal cases: Workday class action and accountability risks. Available at: https://responsibleailabs.ai/knowledge-hub/articles/ai-hiring-bias-legal-cases (Accessed: 28 January 2026).

Cyberhaven (2025) ‘AI insider threats: what security teams need to know’, 18 November. Available at: https://www.cyberhaven.com/blog/insider-threats-in-the-age-of-ai (Accessed: 28 January 2026).

The Guardian (2024) ‘McDonald’s ends AI drive-thru trial’, 17 June. Available at: https://www.theguardian.com/business/article/2024/jun/17/mcdonalds-ends-ai-drive-thru (Accessed: 28 January 2026).

Information Commissioner’s Office (2022) ICO fines Clearview AI Inc over £7.5m and orders UK data deletion, 23 May. Available at: https://ico.org.uk/about-the-ico/media-centre/news-and-blogs/2022/05/ico-fines-clearview-ai-inc-over-75m-and-orders-uk-data-deletion/ (Accessed: 28 January 2026).

Royal Commission into the Robodebt Scheme (2023) Report. Available at: https://robodebt.royalcommission.gov.au/publications/report (Accessed: 28 January 2026).

Economic Times (2025) ‘Company that sacked 700 workers with AI now regrets it — scrambles to rehire as automation goes horribly wrong’, Economic Times, 10 June. Available at: https://cio.economictimes.indiatimes.com/news/artificial-intelligence/company-that-sacked-700-workers-with-ai-now-regrets-it-scrambles-to-rehire-as-automation-goes-horribly-wrong/121734270 (Accessed: 28 January 2026).

CX Dive (2025) ‘Klarna changes its AI tune and again recruits humans for customer service’, CX Dive, 9 May. Available at: https://www.customerexperiencedive.com/news/klarna-reinvests-human-talent-customer-service-AI-chatbot/747586/ (Accessed: 28 January 2026).

Scaling judgement in an age of abundance

Mon, 06 Apr 2026 09:00:00 +0000

In my previous article I explored how AI is shifting the bottleneck in professional work from execution to judgement. Producing things has become fast and cheap, so deciding which of the things produced are actually good has become the real constraint. This article explores what to do about it.

When I use the word judgement here, I broadly mean the work of deciding whether an output is accurate, appropriate, and worth acting on. That includes approval decisions, quality checks, risk assessment, brand fit, client suitability, and strategic trade-offs. What these have in common is that they require a person to apply domain expertise to a specific output in a specific context, and to be accountable for the result. It's that category of work which AI cannot yet reliably do for itself.

The obvious initial response, when any capacity problem appears in an organisation, is to add more people. In many settings that does help: adding reviewers alongside clearer standards and escalation rules can improve throughput and consistency. But the volume of AI output is growing faster than most organisations can hire to keep up with it, and adding more people is at best a partial or temporary answer to a structural problem.

The organisations that handle this well won't hire their way through it alone. They'll design systems for how judgement gets applied and scaled. The end goal will be to build conditions in which AI does a large portion of the checking work, with humans focused on the decisions that genuinely require them. Most organisations have more ground to cover to get there than they realise.

A progression, not a switch

Scaling judgement isn't a binary choice between reviewing everything and reviewing nothing. It's a progression, and understanding each stage is what makes it possible to move through them deliberately.

The first stage is direct review, where a human checks every output before it goes anywhere. This is the right approach at first when you're new to a domain, when stakes are high, or when you don't yet have a reliable picture of how AI performs in your specific workflow. The goal at this stage is learning rather than efficiency, because what you observe here is what makes the next stages possible.

The second stage is sampling, where humans review a proportion of outputs rather than every one, using that sample to monitor quality, catch drift, and adjust when necessary. The logic of this stage transfers from mature quality systems in other industries, but the analogy isn’t perfect because AI outputs can fail in both systematic and sporadic ways simultaneously. A model can fail systematically if prompts, source data, or settings are consistently flawed, and it can also produce ninety-nine accurate responses followed by one highly plausible but completely wrong one, with no visible warning in the surrounding outputs. This makes sampling AI output a more demanding discipline than traditional quality control because the reviewer isn't just spotting an obvious defect. They're spotting a plausible falsehood, and that can require deep domain expertise rather than routine inspection. It's another reason why judgement doesn't scale like execution.

The third stage is designed oversight, where quality standards have been encoded into the workflow itself so that a significant portion of checking happens automatically before output reaches a human reviewer. A second AI pass can evaluate outputs against a defined rubric, flag factual inconsistencies against known sources, or surface potential issues before a human sees the result. More sophisticated implementations assign confidence scores to outputs. These act as priority signals, allowing reviewers to triage their attention and concentrate scrutiny where it's most likely to be needed, rather than treating every output the same. A simpler version of the same idea is a traffic-light system that routes outputs by risk level before anyone opens them.

These mechanisms need careful design because a priority signal is only as reliable as the criteria it measures against and a poorly calibrated system can create false security. They also require real preparation, including stable rubrics for what good output looks like, agreed definitions of quality, and usable source data to check against. For most organisations, that preparation can be a significant blocker in moving to this stage.

Designed oversight can only handle the portion of quality that can be defined clearly enough to check systematically: consistency with guidelines, factual accuracy, compliance requirements. By offloading those routine checks, it frees human attention for the decisions that genuinely require it, the context-dependent calls that depend on deep expertise and are unlikely to be fully encodable. The intention of this stage is to preserve human judgement for the hardest and most important calls.

Calibrating oversight to risk

The progression above isn't a single ladder that an entire organisation climbs together. Different tasks warrant different levels of oversight, and the right question isn't simply "how do we review less?" but "how much human supervision does this particular task actually require, and what are the consequences if the automation fails?"

A task that produces output used directly in a client-facing document carries different risk from one that feeds an internal first draft. A task where an error is immediately visible and easily corrected is different from one where an error might propagate silently through a process before anyone notices. Risk is the calibrating variable, and as execution scales, those calibration decisions need to be made explicitly and systematically.

It's also worth noting that this kind of top-down design only works if the organisation has visibility of where AI is actually being used. Many organisations are still grappling with employees using personal accounts or unsanctioned tools for tasks that carry real business risk, which means a risk framework can only be as good as the picture of execution it's built on. Implementing strong governance procedures can help on both counts, providing the visibility needed to calibrate risk and reducing exposure at the same time.

Organisations already make these risk-based judgements about human-generated work: some documents require a director's sign-off, others just need to be filed. What's new is that AI is increasing the volume of outputs across all risk levels simultaneously, which means those decisions need to be made deliberately rather than left to individual habit.

What this looks like in practice

The clearest public illustration of this progression has played out in software development. In February 2025, Andrej Karpathy, co-founder of OpenAI, coined the term "vibe coding" to describe a way of working with AI where you "fully give in to the vibes, embrace exponentials, and forget that the code even exists." Exactly one year later, he proposed a revised framing: "agentic engineering." His description: "you are not writing the code directly 99% of the time, you are orchestrating agents who do and acting as oversight." The goal, as he put it, is to "claim the leverage from the use of agents but without any compromise on the quality of the software."

That shift from "forget the code exists" to "act as oversight" reflects the progression above. It works particularly well in software because code can be checked against an objective standard by a machine: a compiler tells you whether it runs, a test suite tells you whether it behaves as specified. Software has unusually powerful tools for automated verification, which is why it's ahead of most professional domains in implementing this kind of oversight. The same progression is available elsewhere, but defining what good output looks like requires more work in domains where there's no objective test to fall back on.

Let’s consider marketing as an alternative example. AI makes it easy to produce content at scale with multiple versions of copy, creative variations, audience-specific adaptations, all generated in the time it previously took to produce a single asset. The efficiency gains are real, but they can be risky because studies have found that a significant proportion of consumers say AI-generated advertising makes them less likely to choose a brand. That finding comes with caveats because the effect varies by category, audience, and how visibly AI is involved. The judgement problem in marketing isn't only about whether any individual piece of AI-generated content is good enough, it's also about what happens when the volume of output starts to lower the average quality threshold of what gets published, potentially eroding brand distinctiveness, or reducing the care that audiences notice and respond to. A marketing team that scales its output without scaling the judgement applied to it is accumulating a risk that compounds quietly and is harder to catch than it would be in a domain with objective verification.

Financial analysis presents a different risk profile. The failure modes here are less about tone and more about substance: input data quality, formula logic, modelling assumptions, and whether the narrative interpretation reflects what the numbers actually say rather than what someone hoped they'd say. A finance team that produces more analysis than it can meaningfully check isn't more informed, it's more exposed.

The upstream condition

Both the progression and the risk calibration depend on one thing: being clear about what good output looks like in your domain. Much of what experienced professionals know about quality is tacit, internalised through practice and expressed through judgement calls rather than rulebooks. When standards are hard to articulate clearly enough to encode, that difficulty is worth examining. It can be a sign that the original intent wasn't precise enough, and sharpening it makes the encoding task more tractable. But some high-value judgement is genuinely hard to formalise even when intent is entirely clear, because it depends on context, timing, and the kind of pattern recognition that comes from years of experience in a specific field.

These articles are informed by my experience delivering AI capability programmes with Dixon AI, and by the Purpose, Execution, Judgement framework specifically (see link in the comments). The judgement design challenge is something I see across organisations and sectors, and while the pace and starting point vary, the direction tends to be the same: move from reviewing everything by default, to sampling by design, to building the systems that let human attention focus where it genuinely matters.

AI is shifting the bottleneck from execution to judgement

Sun, 22 Mar 2026 09:00:00 +0000

I spend a lot of my time helping organisations understand how to work effectively with AI. One of the frameworks I find most valuable is the Purpose - Execution - Judgement (PEJ) framework (Dixon, 2026). PEJ proposes that every productive human-AI activity comprises three distinct domains:

Purpose: the why, meaning the goals and direction that determine what you're trying to achieve.
Execution: the how, the doing, creating, and analysing.
Judgement: the which, determining which output actually matters and being accountable for the choice.

I find this framework particularly useful in practice because it gives organisations a clear map of where human value sits in an AI-assisted workflow. Right now, the relative importance of these domains is shifting faster than most people have noticed. While Execution is accelerating, the bottleneck is moving. For most of professional history, that bottleneck was simple: the time it takes to produce things.

Think about what it used to mean to develop a marketing campaign. A team might spend days researching the ideal customer, developing concepts, producing copy and creative, rounds of revision, and sign-off processes. The ideas might come quickly, but the execution took time. The same pattern held almost everywhere. A consultant preparing a client proposal would spend hours writing, structuring, and formatting. A finance team producing a quarterly analysis might take days on the data-gathering and modelling alone. A project manager building a new product: painstaking work, line by careful line. In each case, producing the thing was where the hours went.

Organisations built themselves around this. Teams were sized to match how much they could produce, timelines reflected how long making things actually took, and rising to seniority often meant being fast and good at the work itself. Execution was the scarce resource.

AI is changing this, and it is changing it fast. What used to take a day now takes an hour. What used to take an hour takes minutes. From market analyses to complex project plans, the time cost of production has collapsed, and I frequently see these gains with my clients. But something else is happening alongside it, and it is starting to matter more than the headline efficiency gains.

Let’s return to that marketing team. Previously they might produce a single campaign asset or piece of content in a given cycle. Now, with AI, they can produce variations at scale: multiple versions of copy, creative directions, or audience-specific adaptations in the same amount of time. The increase in output is real and meaningful, but so is the burden it creates, because someone still has to evaluate what is actually worth using.

An even clearer example can be seen in software development. A team working on a new feature might previously have scoped, designed, and built one or two features over a given period. With AI accelerating execution, they can now prototype and build many more features in the same timeframe. The constraint shifts immediately, and the question is no longer "can we build this?" but "is this fully tested and ready for release?" The volume of possible outputs increases, but the capacity to make high-quality decisions about them does not.This is where the new bottleneck sits: not in producing the work, but in deciding what is good, what is relevant, and what is worth acting on.

AI scales execution. It does not scale judgement. Consequently, judgement becomes the new constraint.

This is becoming increasingly visible as AI adoption picks up. A 2026 survey by UnlikelyAI of over a thousand senior decision-makers found that they were spending almost as much time checking AI-generated work as using AI in the first place: around two and a half hours a week on verification, against roughly two hours and forty minutes of use. These are not the numbers of people who have solved the problem of AI evaluation; they are the numbers of people absorbing its cost, one careful hour at a time.

These costs accumulate in ways that rarely show up in productivity measurements. Organisations count the efficiency gains but rarely account for the growing cost of evaluating what has been generated. The result is a gap between adoption and value that the headline numbers obscure. While adoption is up, value is not keeping pace: only 22% of organisations reported significant returns on their AI investments three years after the technology entered mainstream business use. The constraint has shifted, and most organisations have not caught up with where it has moved to.

This is where PEJ becomes directly relevant. In PEJ terms, Execution is the domain AI has transformed. The cost of the "doing" has dropped dramatically and will continue to drop. This exposes Judgement as the immediate downstream consequence of abundant Execution. More output requires more evaluation, and the bottleneck moves.

The professionals and organisations that understand this shift now, and start designing for it deliberately, will have a real advantage.

Effective evaluation of AI output is not a technical skill; it is a domain skill. The capability you need to judge whether an AI-generated concept is right for a specific audience is not knowledge of how the AI works. It is deep knowledge of your field. A machine can generate a hundred versions of a strategy document, but it cannot know whether any of them is right for this client, this moment, or this context. It cannot catch the recommendation that would work on paper but fail given what you know about how the organisation actually operates. Those things need someone who has done the work long enough to develop an internalised sense of what "good" looks like that no brief fully captures.

Expertise does not transfer automatically. Knowing your field well does not mean you will evaluate AI output well by default. That judgement needs to be actively applied through the right habits of scrutiny and, at the organisational level, through the right structures.

The era of effortless production is arriving, but the era of effortless judgement never will.

My next article will explore The Scalable Judgement Model. If judgement is the new bottleneck, how do you design for it at scale?

Why you should be giving AI a role

Thu, 12 Jun 2025 09:00:00 +0000

AI can an incredibly powerful tool, but it's easy to fail to get the best responses out of it if you don't know how to use it. One of the simplest things you can learn to do is to give AI a role whenever you prompt it. AI can do all sorts of tasks, but if you're not giving it a role it won't know how best to respond to your prompt. The difference between “Summarise this” and “You’re a legal expert summarising this for a client” is often the difference between an average response and a brilliant response. Imagine asking a question to an elderly relative and then asking that same question to your manager or an old school friend - the answers would probably be very different.

Let's take a look at an example of this in action. This is the response I received when I asked ChatGPT 'Please give me three suggestions for dinner':

These suggestions are pretty good and I enjoy a thai green curry. However I haven't told it how to respond, so it's taken a guess. Now let's see how the response changes when I simply give the AI a role. My new prompt is 'You are a ten year old child. Please give me three suggestions for dinner':

The AI has now taken on the role of a ten year old and responded appropriately. It's using different language and making suggestions more likely to come from a ten year old, and it's even asking if it can have pudding! Let's explore further and imagine we want something more sophisticated for dinner for a special occasion. We could ask the AI to take on the role of a Michelin star chef:

Now the suggestions are a lot more sophisticated, and they come with much more detail. These suggestions might be ideal for a special meal, but perhaps not for every night!

These examples demonstrate how giving the AI a role can make a big difference to the response, even when the request (Please give three suggestions for dinner) is pretty simple and remains identical each time.

Why roles matter

Generative AI language models don’t actually think, and they're just astonishing mimics. They draw on patterns in the data from across the internet, including job descriptions, tone, and professional jargon. When you tell the AI to act like a teacher, or a recruiter, or a coach, it draws on thousands of examples of how those people communicate and it uses the role you give it to inform how it responds. You're not just giving it a task, you're giving it context too, and this context is what makes answers more relevant. You probably wouldn't ask a designer to help with some data analysis, or a programmer to produce a marketing campaign, because you assign tasks to people with the relevant expert knowledge. Working with AI is similar, but if you don't give the AI a role then the response will be much more generic because it won't know how to respond.

Thinking in roles

Once you get the hang of it, this shift in mindset opens up new possibilities. Instead of just thinking that you're asking a chatbot for help, you might start building an AI team in your head. Need new ideas? Ask a creative director. Need to review your CV? Bring in a hiring manager. Need to challenge your thinking? Appoint a devil’s advocate. The model is the same, but the role you give it changes how it behaves, and what you get from it.

Here are a few suggested roles to try:

Editor: Tightens your writing, suggests improvements, trims waffle
Research assistant: Gathers relevant points, compares viewpoints, finds sources
Coach: Asks questions, reflects back, helps you clarify your thinking
Critic: Challenges your assumptions, spots gaps or weaknesses
Translator: Converts jargon into plain English (or vice versa)
Time-saver: Drafts emails, creates outlines, summarises documents

These aren’t job titles for the AI. They’re lenses: ways to focus the model on what you really need. They help you shape how the model responds, and there are no restrictions on the roles you can assign.

How to craft a great role: adding detail gives better results

To craft a really effective role, don’t just give the AI a job title or position. Add details like the personality or tone you want it to use, or mention relevant qualifications and experience you want it to have. The more detail you give it, the more likely you are to receive a high-quality response to your prompt. Here’s a more advanced example to show how much context I might give to AI for some competitor market research:

You are a Market Researcher, a commercially minded investigator responsible for identifying, analysing, and comparing key competitors within a defined market. Your role is to uncover strategic insights that inform business decisions by mapping out the competitive landscape with precision and clarity. You begin by identifying the most relevant competitors based on market positioning, customer base, product relevance, and geographic or sector overlap. You gather detailed information about each competitor’s products and services, including features, pricing models, target customers, delivery methods, and innovation levels. Your analysis goes beyond description. You evaluate the strengths and weaknesses of each competitor’s offering by comparing factors such as value proposition, market share, customer perception, technology use, differentiation, customer service, and marketing strategy. You pay close attention to emerging trends, gaps in the market, and areas where competitors may have a strategic advantage or vulnerability. Your work is structured, clear, and insight-led. It is designed to support decision makers in product development, marketing, and strategy. You synthesise findings into actionable intelligence, highlighting opportunities for differentiation, innovation, or repositioning.

There is no limit to the amount of detail you can give, and don't forget, you could even use AI to help you write the role description!

The Red Giant says...

Using AI well isn’t just about knowing what to ask. It’s about imagining who you want to ask, and who you want on your team. The better you are at assigning roles, the better the model becomes at helping you.

In the future, we may not just prompt AI. We may manage it, assembling flexible teams of digital collaborators who help us think, create, and grow. If you're not already using this technique then start by giving the AI a role next time you use it, and once you get used to it you'll be surprised at how much difference it can make.

Bias in the machine: why fairness in AI is everyone's problem

Thu, 05 Jun 2025 09:00:00 +0000

Artificial Intelligence should promise impartiality. Unlike humans, AI doesn’t get tired, impatient or hangry, and AI outputs are purely based on the inputs, meaning AI disregards characteristics such as gender, accent, or skin colour of the person providing the input. And yet, over and over again, it turns out AI is regularly biased, and sometimes spectacularly so. AI systems can cause significant, often unintentional, negative impacts. They might generate only male images for CEOs and only female images for nurses, or they might offer higher-interest credit loans to minority groups. In a well-publicised example at Amazon, an AI recruitment system downgraded women’s applications because they didn’t look like the ones that succeeded in the past. The system had taught itself to prefer male candidates, so it penalised terms like “women’s” and downgraded graduates of all-female colleges. Amazon tried to fix it but ultimately they scrapped the tool because it couldn’t be trusted to make fair decisions.

So what’s going on here? And more importantly, what can we do about it?

What is bias in AI, and why does it matter?

Bias in AI occurs when systems make decisions that are unfair, prejudiced, or systematically skewed, often because of the data they were trained on. AI models learn from data, and as the saying goes "garbage in, garbage out". If AI models are trained on biased data from the internet, they are likely to reflect back the world as the internet portrays it, a world full of historical inequalities and structural imbalances.

This matters to everyone because AI is increasingly involved in decisions that affect people’s lives, in hiring, education, healthcare, finance, and even law enforcement. Even small amounts of bias in these systems can scale quickly and quietly. Importantly, people may not even know they’ve been treated unfairly or have any way to appeal. The results aren’t just embarrassing, and they can be damaging. Recent examples have shown bias in mortgage applications, healthcare and recruitment.

How does bias creep into the system?

There are three common culprits:

Biased data: Most modern AI models are trained on enormous datasets scraped from the internet or drawn from historical records. If the source material is biased, as much of the internet is, the outputs will be too.
Poorly defined goals: If an AI system is trained to optimise for the “best” candidate or the “most relevant” image without very carefully defining what that means, the AI system will make its own assumptions. Often the outputs will then reflect existing stereotypes.
Lack of oversight: AI models are often developed behind closed doors with little transparency. Without rigorous testing and visibility of the training data, biases can go unnoticed until they cause real and noticeable harm.

Can’t we just tidy up the data?

Cleaning up data helps, but it’s not enough. Bias doesn't just lie within the data and it can also occur in how we frame problems, define success, and interpret outputs. Even seemingly neutral inputs can produce skewed results when filtered through a flawed model or incentive structure.

Debiasing AI isn’t a one-time data-cleaning job. It’s an ongoing process that includes:

Diverse teams building and testing models
Clear standards for fairness and accountability
Transparency about how systems work and are evaluated
Input from affected communities, not just engineers

Right now, most major AI systems are what experts call “black boxes.” Their inner workings are secret, and their impact is often hard to trace. That’s a problem, particularly when they’re used in public services or critical decision-making.

What can you do about it?

If you’re building or using AI tools, whether in education, hiring, healthcare, or beyond, here are some practical steps you can take:

Ask questions about training data. Where does it come from? Who is represented, and who isn’t? Where might bias occur?
Test for edge cases. Does the system behave differently across different demographic groups? Try a few tests with different backgrounds, different genders or different names to see if and how the output is affected.
Monitor outputs regularly. Bias isn’t always obvious at first glance, so look for patterns in the outputs over time.
Push for transparency. If you’re buying or using AI tools, ask how they were tested and what measures are in place for fairness.
Invest in AI Literacy. AI literacy isn’t just about how to prompt, and it's also important to learn to critically evaluate the outputs and spot when something seems off.

Using AI responsibly doesn’t mean avoiding it altogether, and it requires adopting AI with eyes open so you can understand how the tools work and evaluate the outputs.

The Red Giant says...

Bias in AI isn’t just a bug, and its outputs often represent a mirror held up to society. That means we can do something about it. Building fairer AI starts with recognising that bias exists, asking questions, and making deliberate choices about how these tools are trained, tested, and used. We can’t code our way out of inequality, but with care and accountability and responsible use, we can make sure that the AI systems and tools we build reflect our best values and not our worst habits.

Is AI going to take my job?

Thu, 29 May 2025 09:00:00 +0000

How AI is already impacting jobs

There's a saying that's been doing the rounds recently along the lines of 'AI won't take your job, but somebody using AI probably will'. How true is that, and what does it mean for you? On one hand, Anthropic CEO Dario Amodei claims that AI could eliminate 50% of entry-level office jobs within 5 years, but on the other hand, many companies aren't yet seeing a significant impact from AI. Let's take a look at a couple of examples to see what's happening.

Earlier this year, Duolingo announced an 'AI-first' strategy. This strategy included using AI to create more language courses, replace contract workers with AI, and telling teams that headcount would only be given if they could not automate more of their work. The announcement and implementation of this AI-first strategy led to significant backlash online, with users expressing concerns over potential job losses and the quality of AI-generated content. Recently CEO Luis von Ahn acknowledged mistakes he made when he announced the AI-first strategy:

One of the most important things leaders can do is provide clarity. When I released my AI memo a few weeks ago, I didn’t do that well.

It's reassuring to see a CEO show public humility and to address his mistakes. It's also important to note that he hasn't stepped away from the AI-first strategy. What he's done is address the implementation and he has clarified that AI is intended to enhance, not replace, human work. Importantly, he emphasised the company's commitment to supporting its employees through the transition.

Now let's also look at the case of Klarna, the Swedish fintech company. Last year Klarna became the first European company to launch a ChatGPT plug-in and they replaced 700 customer service employees with AI-powered chatbots. Initially, this shift was touted as a big success, with the AI handling two-thirds of customer service interactions and performing the work of 700 full-time staff. However, the company has since faced challenges, including customer dissatisfaction and quality issues. Recognising these shortcomings, Klarna recently began rehiring human staff to complement its AI systems, aiming for a more balanced approach. Klarna CEO Sebastian Siemiatkowski explains:

I just think it’s so critical that you are clear to your customer that there will be always a human if you want

Klarna haven't completely backtracked and they're not rehiring as many people as they replaced because they're still using AI for many of their customer service interactions. They probably went too far, too quickly, and they suffered as a result. Klarna is now shifting to a more balanced approach of humans working alongside AI.

These examples highlight a critical point: AI is already reshaping the workforce, but its implementation is complex. The shift to large-scale AI implementation is happening at different speeds for different industries, and many organisations might not get it right first time. For organisations considering AI implementation, starting small and building from initial wins often work best. While AI can enhance efficiency and productivity, it also raises questions about job displacement, quality control, and the need for human oversight. The key takeaway is that AI's impact on jobs will be nuanced. In the long run I'm convinced there will be a significant shift for almost all organisations, but it requires thoughtful integration and a focus on complementing human skills rather than outright replacement.

Working with AI

For all the headlines about AI replacing workers, many of the most meaningful changes are happening in more subtle ways. Rather than wholesale replacement, we’re seeing the rise of human-in-the-loop processes where people and AI work together, each focusing what they do best. For instance, an expert may carefully define the task, using their domain expertise and industry knowledge, use AI to help actually implement the task, and the expert may then review the output to ensure accuracy and quality. This final check is crucial, and it shouldn't be skipped.

This method has been seen to work well in a wide range of tasks. Customer support teams now use AI to automate simple responses, but in more complex responses the AI can be set to draft replies that humans edit or approve prior to sending. Marketing professionals are turning to AI tools to generate campaign ideas which they then finalise, while designers and coders increasingly rely on AI assistants to speed up their work. In many cases, AI acts like an extra team member: tireless, fast and helpful, but still in need of direction.

We’re also starting to see the rise of AI agents, systems that can take on entire multi-step tasks with minimal supervision. These agents are good at automating repetitive processes or coordinating digital workflows, from booking meetings to pulling reports. However, even the most advanced agents still need oversight, and people remain essential by setting objectives, reviewing outputs, and making the kind of nuanced decisions machines can’t.

So while some jobs are being reshaped, others are simply evolving at a slower pace. Roles that involve judgment, empathy, creativity or complex coordination are proving harder to automate and they often benefit the most from AI support. In this new working relationship with AI, humans aren’t being replaced, but they are being upgraded and expectations will rise.

Staying Ahead of the Curve

So what can you do to stay valuable in an AI-shaped job market? You don’t need to become a machine learning expert, but I'll keep emphasising the importance of becoming AI literate

Look for ways to make AI complementary to your work. That involves leaning into your human strengths, whether that's decision-making, problem-solving, communication, ethical judgment, or adaptability. These are harder to automate and increasingly valuable in teams where AI is starting to do more of the heavy lifting.

It also helps to understand how AI fits into your industry. In healthcare, it might be diagnostics and data, whereas in finance, it could be analysis and fraud detection. In education, AI presents huge opportunities for personalised content and student support. The more you know about where AI is headed in your field, the better you can steer your career alongside it and look for suitable opportunities.

One way to think of AI's likely impact on work is this: the internet didn’t replace everyone who didn’t learn to code, but it did quickly reward the people who figured out how to work online, and now most work is online. AI will probably be similar. There will be significant advantages for organisations who successfully implement AI first, and then within a few years we'll all probably need to learn to work with AI.

The Red Giant Says...

AI isn’t yet coming for everyone’s job, but it is already changing how most of us work. This transformation is not happening all at once. Some industries are feeling the change already, while for others it may not be noticeable for a while. But the idea that “AI won’t take your job, someone using AI will” is becoming more true by the day.

If you haven’t started using AI regularly in your work, then it's a good idea to start sooner rather than later. And if you don't know where to start, find somebody in the know to help you. Ignoring AI could end up being like ignoring the internet in the early 2000s, or saying no to computers in the 90s, and it isn't a great long-term strategy.

It’s true that AI will probably completely replace some jobs, and it will probably reduce the need for people in others. But history also tells us something worth remembering: big technology shifts usually create more jobs than they destroy. If we can learn to work with AI, rather than compete against it, then there’s likely to be plenty of work to go around for a good while yet.

Coming next...

In the next issue of The Red Giant I'll be looking at bias in AI systems, and how we can mitigate the effects of bias.

Prompting: unlocking the magic of AI

Thu, 22 May 2025 09:00:00 +0000

What Is Prompting And Why Does It Matter?

Do you ever use AI to generate content or answer questions, only to find that you're not getting the same high-quality results as other people? Or perhaps you find the quality of outputs is inconsistent? If so, learning some basic prompting techniques could be what you need in order to unlock the power of AI. Before we get into prompting techniques and some additional practical tips, let's take a look at what prompting is.

Put simply, prompting is the way we tell an AI what we want. The good news is you don’t need technical abilities to be a good prompter, and basic techniques are easy to learn. A well-written prompt is like giving instructions to a helpful colleague: the more information you provide and the more specific you are with your request, the more likely you are to receive a useful response. If you don't give AI enough detail, it has to make up the missing detail. You may get lucky and it could give a good response, but without providing that detail you are leaving it to chance.

Let's take an imaginary example of a 6 year old child asking you to explain electricity. You decide to use AI to provide the explanation, and here are two prompts you could use:

1) Explain how electricity works
2) Explain how electricity works to a 6 year old child

If you use the first prompt you're likely to find that it gives you a long and quite technical response, which is unlikely to be appropriate in this scenario. The AI is unaware of who it's providing the explanation for, so it takes a guess. In the second prompt, you've added context by telling the AI that you want an explanation for a 6 year old, so it is much more likely to give a suitable response. You could go further and add additional details, or you could leave it like that. The amount of detail you need to add depends on the scenario, and there are no rules, so you can include as little or as much as you choose. Try modifying the above prompts by adding phrases like ‘make it funny’ or ‘use cats as an analogy’ to see how the output changes.

Once you get the hang of it, prompting quickly becomes less of a guessing game or something you need to remember, and pretty soon you'll barely need to think about it when you use AI. Whatever you choose to use AI for, whether you’re drafting content, analysing data, brainstorming ideas or automating tasks, a good prompt makes all the difference.

RTF - A Simple Prompting Technique That Works

There are many different prompting formats that have shown to produce good results, and as you get more experienced with prompting you can even mix-and-match prompting techniques or make up your own to fit your specific request. A basic format I recommend learning is role-task-format (RTF), which will usually provide the AI with enough information to give a decent response. Let's take a look at how it works.

Role: Assign the AI a role so it knows how to respond effectively. AI has a huge amount of knowledge and capability, so it can take on almost any role you can think of. For example, you could ask it to be a successful entrepreneur if you want to ask it about business ideas, or you could ask it to be an award-winning author if you want it to assist with creative writing. You can keep the role simple, or you can choose to include a lot of detail if you want. For example, you may choose to list specific areas of expertise or qualifications you want the AI to have, or you could even describe what type of personality it has!
Task: Describe the task you want the AI to do. The more detail you give it and the more specific you are with your request, the more you can control the output. For example, if you ask it to do a competitor analysis, you could specify whether you want it to focus on pricing, marketing strategies, or product features, or how many competitors to include. Alternatively, if you're asking AI to write a blog post, you might choose to give it brief details about what to include or suggested section headers.
Format: Describe the format you want the AI to provide the output in. Unless you request something else, AI will usually provide a text output. You can control the output by giving details on what you'd like. For example, you could give a word limit, or you could specify which language to use. You can also specify whether it uses bullet points, write in prose, or even outputs the response as a rhyming poem! The response doesn't have to be text, and you can ask for alternative formats, for example, asking for a table in a spreadsheet, or lines of Python code.

Here's an example of a complete prompt using this role-task-format structure.

You are an award-winning technology journalist with experience writing for non-technical audiences. Write a 600-word blog post in British English explaining the basics of quantum computing in simple terms, using relatable analogies. Present the information in three clear sections with subheadings and a short summary at the end.

Learning this basic technique will significantly improve the quality of your results, and there are plenty of other techniques you can learn too. For example, you could choose to add background to an RTF prompt to provide the AI with further background detail about a specific company or person relative to your prompt. There are many other techniques, and you could also include step-by-step instructions, examples, details of scope or other context as needed. A good prompt can end up being very long, potentially even thousands of words long, and the more detail you give the AI, the better quality the response is likely to be.

Some further practical tips

Getting comfortable with basic prompting techniques is a great start, and here are some additional simple techniques that can save you a lot of time and really help the AI give you what you're looking for:

Provide an example: If you share with the AI an example of what you want then it's much more likely to reply in a similar format. For example, if you're using AI to write a report, share an example of a similar type of report. This technique works for images too if you want it to produce an image in a certain style.
Ask AI for help with prompting: If you're not sure how to write a good prompt for a particular task, AI can often provide you with a good prompt. Remember to give the AI a role when you ask it for a prompt, so in this case you could ask it to be an expert prompt engineer!
Ask AI what information it needs from you in order to complete the task to the best of its abilities: This is a great little tip to ensure that the AI has all of the relevant information from you and you haven't forgotten anything in your prompt.
Re-use your prompts: Good prompts can be very long and detailed, so I suggest getting in the habit of saving and reusing prompts. Share successful prompts with colleagues, and avoid starting from scratch each time you need a detailed prompt.
Tell the AI to use code: Generative AI models can be bad at maths, but they have access to tools which can help them. If you're asking a question involving maths or coding then adding the simple instruction for the AI to use code will significantly increase the chances of getting the correct result.

I hope you find these tips useful, and please share with me any additional tips you may have so that I can collate them and share them in a future article.

The Red Giant says

Prompting isn’t about learning a complex language and it’s just about being clear and specific. The more context and direction you give to the AI, the better the results you’ll get. There are techniques to explore, but you don’t need to master them all at once. Start simple by adding more detail to your requests and specifying the format you want back, and you’ll quickly notice your prompts improving.

Coming next

In the next article of The Red Giant I'll begin to explore how AI is reshaping the jobs market and what this could mean for you.

Learning and assessment in the age of AI

Thu, 15 May 2025 09:00:00 +0000

What AI Means For Learning and Assessments

The integration of AI in education has sparked a lot of debate with some educators seeing AI as a catalyst for much-needed system reform, while others fear it threatens academic integrity and cognitive development. Both perspectives have merit, and outcomes will largely depend on the quality of the AI implementation. With thoughtful application and investment in AI literacy, AI has the potential to transform teaching, learning, and assessment practices. However, navigating this transformation successfully requires acknowledging both the opportunities and the challenges, and poor AI usage is likely to have a negative impact on learning. AI should not be used everywhere and there remains a vital place for traditional teaching and learning methods, so knowing when and where to use AI is crucial in helping educators maximise their students' potential.

In this article I'll be exploring some of what's already happening with AI in education and what that could mean for the future of learning and assessments. Let's start with some reviewing some facts on where we are today:

Students are already using AI: Research shows that students are one of the largest groups of AI users, and the indicators are that student usage is likely to keep growing as use cases and AI tools evolve. The Kortext study I've mentioned in a previous article found that 92% of UK students are already using AI in some form, with 88% having used generative AI for assessments. It's important for educators at all levels to understand how students are using AI. In Higher Education specifically, the excellent Anthropic Education Report also provides detailed insights into how university students are using Claude. Both Claude and ChatGPT have recently offered free usage to students in the US, so I expect we'll see many more detailed higher education insights over the coming months, along with additional insights on AI usage at other levels of education.
Investment in staff development is still low: The recent AI and Jisc's Leadership Survey 2025 reports that staff development in AI remains low despite AI being cited as a major challenge across both Further and Higher Education. It's reassuring to see the increases in AI activities since last year, but AI literacy is crucial for all staff in order for an institution to thrive. Working groups, policies and guidance are a good start, but staff also need a structured approach to getting hands-on experience with generative AI tools. The survey shows fewer than 40% of HE respondents reporting that their organisations have conducted staff development on AI, so this remains a major area for improvement.

Source: AI and Jisc's Leadership Survey 2025 - Artificial intelligence
Advanced AI models can solve pretty much any assessment without being detected: Although poor use of AI is relatively easy to detect, students with strong AI literacy can use AI to solve pretty much any traditional assessment, from writing an essay, business case or marketing plan to solving complex maths, coding and even creative assignments. Achieving these results requires understanding which AI model(s) to use - the paid version of ChatGPT currently has seven to choose from! - good prompting abilities and an understanding of the subject in order to review the output carefully. Educators really need to be aware of the capabilities of the latest AI tools when designing assessments to ensure that assessments are truly testing their intended learning outcomes and not just how well a student can use AI.
Research shows using AI can enhance cognition: This recent article published in Nature analyses 51 studies published since ChatGPT launched in late 2022. The results show that ChatGPT has a large positive impact on improving learning performance along with a moderately positive impact on enhancing learning perception and fostering higher-order thinking. As these tools continue to improve, and as the AI literacy of students also improves, it seems reasonable to suggest AI may have an even greater positive impact in the future. However this impact is unlikely to be universal and AI is perhaps best seen as a multiplier, further enhancing the capabilities of the strongest students and those with access to the greatest tools, while potentially weakening other students. Educators and institutions need to carefully monitor AI's impact on different student demographics and manage this potential attainment gap (through initiatives such as providing access to premium AI tools and support, and carefully designed activities and assessments) to ensure educational AI usage is not promoted in a way which increases inequality.

Where To Start with AI in Education

Unfortunately there is no single 'silver bullet' or 'one-size-fits-all' answer to the question of how to assess in the world of AI, but there are some clearly defined actions which can help educators prepare better assessments:

Develop AI literacy: Learn the basics of AI literacy, and then familiarise yourself with both good and bad practices of AI usage in education
Get hands-on experience: Practice use the AI tools that students are using to understand their capabilities
Collaborate with colleagues: Share findings and questions about AI implementation, don't be afraid to ask questions or admit what you don't know. AI is still relatively new and everybody is learning
Establish an information routine: Find a sustainable way to keep up with major AI developments
Commit to continuous learning: Recognise that regular upskilling is becoming essential

For those working in Higher Education looking for practical resources, here are some options I highly recommend:

Take either (or both) of these short online courses: AI in Education online course on FutureLearn or the AI in Education: Leveraging ChatGPT for Teaching online course on Coursera. Both are excellent, suitable for all, and provide you with practical use cases. The FutureLearn course, which is free for three weeks from when you start the course, has recently been fully updated and led by Martin Compton, the AI and Innovation in Education Lead for King's College London. The Coursera course, which I've recommended previously, is also free and led by Ethan Mollick, professor and Co-Director of the Generative AI Lab at Wharton
Review this Manifesto for the Essay in the Age of AI, a project led by Claire Gordon at London School of Economics and Martin Compton (again) from King's College London, as a great example of how a wide group of academics came together to explore the future of the essay in the age of AI
Read the excellent article Developing a Conceptual Framework for Artificial Intelligence (AI) Literacy in Higher Education, which has been nominated for the 2025 Article of the Year Award by the Journal of Learning Development in Higher Education. It's provides a simple to follow framework along with practical advice on how to implement it.

I'm sure there are many more high quality and freely available resources, so please let me know if you've got any other recommendations and then I'll collate a longer list of resources to share in a future article.

The Red Giant says...

AI is changing how we assess learning and that change can feel unsettling, but discomfort often signals a forthcoming opportunity. With the right training and a dose of creativity, I strongly believe that educators can design assessments that are fairer, richer and more meaningful. There will not be a one-size-fits-all fix and the solutions for a literature course, a coding course, a business plan and a chemistry practical are all very different. But what is universal is the need for AI literacy. The better educators understand these tools, the better they will be at creating engaging learning, and crafting assessments that truly reflect what learners know.

Coming next

In the next issue of The Red Giant I'll begin to look at AI prompt writing techniques, one of the key foundations of AI literacy, and how better prompting is a quick way to lead to better results.

Not (yet) seeing an AI boost? You're not alone

Thu, 08 May 2025 09:00:00 +0000

The Reality Behind The Hype

You may already be using AI to draft emails and documents, generate ideas, or summarise meetings, and if you're seeing notable productivity gains from these tasks, then congratulations because you're ahead of the curve! But don't worry if you're not yet seeing notable productivity gains from your AI use because you're not alone.

The Hype: Recent years have seen headlines proclaiming AI as a transformative force poised to revolutionise every aspect of work. Headlines and reports have declared that AI can 10x productivity, AI will write 90-100% of code in future, and as many as 30% of jobs could be automated by 2030. If these numbers are to be believed, shouldn't you be seeing more significant gains when you use AI?

The Reality: Most people I've spoken with haven't yet seen noteworthy gains from using AI, and I frequently speak to people spending as much time editing and re-editing AI outputs as they would have spent writing those outputs from scratch. Failing to see an AI productivity gain is common, particularly when people haven't been properly trained on how to get the best out of their AI tools or when workflows haven't evolved to take AI tools into account. This lack of significant productivity gain shows in the data, and although reports vary on exactly how much time people are saving with AI, the average AI productivity gain in early 2025 seems to be only around 7% or 2-3 hours per week. This average is increasing and up on 2024, but it's still small compared to where it could be.

You've made a great start if you're already regularly engaging with AI tools, and the reality is that most people haven't yet experienced the game-changing productivity boost that the media touts. So what's behind this gap between the media narrative and real-world usage? This discrepancy isn't due to a lack of effort or desire. It's a reflection of the current stage of AI literacy and AI integration into the workplace.

Many organisations are still exploring how best to implement these tools effectively and responsibly. Most medium-large organisations already have successful pockets of AI implementation, but those AI capabilities are rarely spread effectively across entire organisations. The Boston Consulting Group 2025 AI Radar survey reported that 71% of organisations had trained fewer than 25% of their workforce on AI, so there is a lot of training to do on upskilling employees. There are also many AI policies to write, workflows and processes to be redesigned, and large volumes of data to be cleaned if high levels of productivity gains are to be realised.

This may sound like a lot of work and therefore the pace of successful AI implementation will be varied, but the rewards will be significant for organisations and individuals that move quickly and get this right.

Where The Gains Are Already Happening

While the average productivity boost from AI remains modest, some industries and roles are already experiencing headline-grabbing gains. Understanding these success stories provides valuable insights into where AI currently delivers the most value, and why these cases are currently exceptions rather than the rule.

Software development and coding already show significant benefits of AI implementation. Of the 65,000+ developers who responded to the Stack Overflow 2024 Developer Survey, 81% agree increasing productivity is the biggest benefit of AI tools and I'm sure that number will have increased since then. Reassuringly, 70% of those professional developers do not perceive AI as a threat to their job, and rather they see it as a tool that will increase their productivity.
Content creation stands out as another clear segment where AI can provide significant productivity gains, with copywriters, marketers, and communications specialists reporting significant time savings. Recent reports from Zebracat and Jasper's 2025 State of AI in Marketing giving impressive statistics on these gains. It's important to retain your voice with AI-generated content, so I don't suggest using AI to fully produce content. However, it can quickly get you 80-90% of the way if you have good prompting techniques, and you should always ensure a thorough human review before the output is finalised.
Customer support has also seen meaningful improvements through AI-powered triaging and response suggestions. Zendesk's 2025 CX Trends report shows 79% of customer experience agents believe having AI as a co-pilot increases their abilities, enabling them to deliver superior customer service. Again, the key distinction in successful implementations is that AI augments human agents rather than replacing them entirely.

What separates these success stories from areas still struggling? Three factors consistently emerge:

Well-defined scope: The most successful implementations focus AI on specific, clearly defined tasks rather than broad responsibilities
High-quality training data: Teams with structured, accessible historical data see better results than those with fragmented or inconsistent information
Reimagined workflows: Rather than simply inserting AI into existing processes, the biggest gains come from rethinking workflows around AI capabilities

These patterns suggest that while general productivity gains remain modest, targeted AI applications with appropriate expectations and infrastructure can deliver transformative results. The challenge facing most organisations isn't the AI technology itself, but the people changes required to fully leverage AI.

Strive For Progress, Not Perfection

If you're looking to achieve meaningful AI productivity gains, then I suggest developing a consistent practice of small improvements rather than searching for the perfect tool or perfect prompt. Think of it as compound interest for your workflow; many smaller gains that add up dramatically over time.

One successful technique is to start by identifying those repetitive tasks that drain your time and don't leverage your unique skills. For many knowledge workers, this includes drafting routine communications, organising information, and performing initial research. These areas typically offer the lowest-hanging fruit for AI productivity gains.

Then review that task list and pick a task or process to start where you think AI may be able to help you. Spend 10-15 minutes coming up with ideas on at how AI can help you save time with that task - remember you could even use AI to help you with suggestions here! Then give it a go, and next time you need to do that task, try using AI too. It doesn't matter if you don't get the prompt perfect first time around, and the key thing to start with is experimenting and looking for small wins. This practice builds confidence and helps identify patterns where AI works best for your specific needs, and over time you'll refine your prompts and increase those productivity savings. Share what you find with colleagues, and encourage others to share too. Even saving just 15 minutes per working day translates to more than 60 hours annually - well over a full working week!

Remember that the most productive AI users aren't those with perfect prompting skills - they're the ones who consistently experiment, refine their approach, and integrate these tools into their daily workflow.

From Playing to Building

If you're a more experienced generative AI user you might be stuck in what we could call the 'prompt plateau', that stage where you've mastered basic AI interactions and have become good at prompting, but haven't yet harnessed AI's deeper capabilities. Moving beyond this plateau requires shifting your mindset from simply using AI tools to actively building with them.

The first step is developing a systematic approach to prompting. Rather than crafting one-off prompts for each task, create reusable templates for common workflows. For instance, develop a standard prompting template for content reviews, data analysis, or creative brainstorming. This approach not only saves time but leads to more consistent results.

Next, focus on stringing AI interactions together to solve multi-step problems. Instead of viewing each AI use as isolated interactions, consider how outputs from one prompt can feed into another. For example, you could use one prompt to organise information, another to analyse it, and a third to communicate findings. This approach lets you tackle more complex challenges that a single prompt may not address effectively.

For those of you with some technical background, bigger breakthroughs come from integrating AI tools more deeply into your workflows and looking at agents or automations. Even basic integrations, like automatically feeding meeting transcripts to your AI for summaries or connecting document repositories for context-aware assistance, can dramatically increase your productivity. You could also consider building small custom tools that combine AI with your organisation’s specific knowledge and data. Alternatively, platforms like Zapier, Make, and Microsoft Power Automate allow you to create automated workflows without writing code, and all of them now offer AI integrations that are accessible to non-developers.

Remember that the most valuable AI implementations don't simply automate existing work and instead they enable entirely new possibilities. Ask not just 'How can AI help me do my current work faster?' but 'What else could I accomplish if certain tasks were automated completely?'. The path from playing with AI to building with it isn't about technical complexity, but it's about developing an approach that integrates AI as a component in your broader workflow rather than using it as a standalone tool.

The Red Giant says...

Don’t worry if you’re not yet seeing huge productivity gains from AI. Most people aren’t. If you’re using AI regularly and starting to see small improvements, you’re already ahead of the curve. Keep experimenting, keep iterating, and remember that small gains can significantly add up over time. And don't forget you can use even AI to help if you're not sure where to look for your next productivity gain.

In my experience, wins often come when you convert 'unknown unknowns' (where you're unaware of what you don't know) into 'known unknowns' (where you become aware of what you don't know). For example, moving from 'I didn't know AI could help with meeting summaries' to 'I know AI can summarise meetings, but I'm not sure how to implement it effectively.' This comes from experimentation, talking about AI, and sharing results. Once you've reached the stage of known unknowns, then you can make much better informed decisions about your next steps.

Coming next

Next time in The Red Giant I'll be doing a deeper dive into some specific areas of AI use in education, looking at how AI is impacting assessments and what that might mean for the future of assessments.

Why everyone needs AI literacy

Thu, 01 May 2025 09:00:00 +0000

Everyone needs to learn AI literacy, just as we all had to learn how to use the internet or a mobile phone. AI is already part of our everyday lives, yet most people underestimate just how significantly it will shape their lives in the years ahead. Whether you’re working, studying, job hunting or retired, a basic understanding of what AI is (and what it can and can’t do) is fast becoming essential. AI is often misunderstood, misused, and overhyped in the media. For instance, headlines often claim AI will 'replace all jobs' in certain industries which oversimplifies reality. So what really is AI, and why should you care about being AI literate?

What is Artificial Intelligence?

Let’s start with the basics. Artificial Intelligence, or AI, is a broad term for machines that can mimic certain aspects of human intelligence, usually by spotting patterns in data and making predictions. Most of the AI we use today isn’t conscious (although there is starting to be some debate about that which is a topic for another day) and it's more like a supercharged calculator with a lot of training.

There are two main types of AI you need to know about:

Analytical or Traditional AI, often called machine learning, powers many of the systems we interact with every day. When your email filters out spam, that’s AI. When your phone suggests the next word while you’re texting, that’s also AI. AI is behind Spotify and Netflix recommendations, voice recognition on your smart speaker, the route suggestions in your maps app, and even the fraud alerts from your bank. These tools are trained on vast amounts of data and are good at doing one particular task, such as spotting patterns, making predictions, classifying data, or flagging outliers. This type of AI has been around for many years, and is mostly confined to more technical roles, although the use cases are broadening and will continue to do so.
Generative AI creates new content based on patterns it has learned. Tools like ChatGPT, Gemini, Midjourney, Suno and Claude which can generate text, images, music, and code fall into this category. They use the data they have seen during training to produce an original output. The key thing to understand is that these systems predict what is likely to come next in a sequence when they generate their outputs, whether that is a word, a line of code, or part of an image. This works surprisingly well in many cases, but it also means they sometimes produce false or misleading information, which is known as a hallucination. Generative AI has only really been around in the public domain since the launch of ChatGPT in late 2022 and when you hear hype about AI writing novels, creating art, coding entire applications, or replacing creative professionals, it's most likely that people are referring to generative AI.

You don’t need to be an AI expert to use AI, but knowing the basics, like the difference between analytical and generative AI, helps you ask better questions and make more informed choices about if and when you should use AI. Once you move past the jargon and hype, AI becomes less intimidating, far more interesting and most importantly, far more useful.

Why do I need AI literacy?

Thankfully becoming AI literate is relatively simple. Here’s why it's important for everyone:

Work and learning are changing fast: AI is already reshaping how jobs are done, how people learn, and how we adapt to new tools. Whether you’re starting out, mid-career, or trying to keep up with the latest changes in your field, understanding how AI works helps you stay relevant and confident, rather than overwhelmed or left behind. Students are the largest demographic of AI users and they'll be joining the workforce soon, so companies are adapting their hiring to attract the brightest talents. Duolingo and Shopify have recently announced AI-first strategies that increasingly value AI literacy at all levels of their organisations, and I expect many more companies to follow.
It helps you spot when something isn't right: AI-generated content isn't always reliable. A concerning example is deepfakes - manipulated videos or images that use AI to show people saying or doing things they never actually did - which can range from celebrity face-swaps in movies to dangerous misinformation like fake political speeches or fabricated evidence. As AI tools become more accessible, being able to identify potential deepfakes is becoming an essential digital literacy skill, just like recognizing phishing emails. Sometimes AI-generated content is wrong intentionally as a deepfake, but sometimes it's biased, and sometimes it just sounds convincing while being completely made up. AI literacy helps you develop a sharper eye and equips you to ask better questions about the content you see.
It helps you make smarter, more responsible choices: Being AI literate means knowing when to use AI and how to use it wisely. You’ll understand how AI works, know what it can do, when to trust its outputs, when to double-check, and when a human decision is still the better call.
It helps the next generation: Young people are growing up surrounded by AI in their apps, their games, and even their schoolwork. As parents, educators and role models we have a responsibility to understand these tools, so we can offer guidance, ask the right questions, and model responsible use.
It unlocks new kinds of creativity: AI isn’t just about speed or efficiency and it can also be a creative partner, helping you write, design, brainstorm or build. But to use it well, you need to understand how it works, and where human intervention, oversight and collaboration are required. People argue that AI can't be creative, however when it's used as a collaborative tool, AI can enhance your creativity.

How to start developing your AI Literacy

Now we've discussed why it's important, let's dive into how to develop your AI literacy. There are already many training programs going on across organisations and even across entire countries to upskill on AI Literacy, so developing AI literacy isn't just your personal responsibility. Both the US and Chinese governments have recently announced huge investments in AI upskilling, including programs to embed AI literacy in education from a young age, and I'd expect the EU and UK to follow. Your workplace or place of study may already offer an AI literacy programme, but if not, I strongly encourage you to start developing these skills yourself. The good news is that you don’t need to take a course or read a textbook to start learning how to use AI. The best way to learn is by doing, so please ensure that whatever training you do includes actually using the tools. The time you invest in learning these tools is likely to be repaid many times over, and I also expect you'll have some fun as you learn. Here are four simple ways to get started:

Get hands-on and curious: The simplest and most effective way to learn about AI is to use it. Try tools like ChatGPT, Gemini or Claude and see what they can do, and just as importantly, what they can’t do. They all have free versions which are more than good enough to get started. Here's an example to try: Open an AI tool and type 'You are an AI expert, and I'm learning about AI literacy. Explain three ways AI might help me in my daily life, and one limitation I should be aware of.' Then try modifying your prompt to see how the response changes. The more time you spend experimenting, the more comfortable and confident you will become. Start with small, everyday tasks like drafting an email, summarising a news article, or planning a trip, and get used to how the tool responds so you can begin to find uses for it in your daily life.
Learn the basics of prompting: How you ask (known as 'prompting') matters with AI and there are many techniques for effective prompting. One powerful and basic technique I recommend to start with is the Role - Task - Format (RTF) approach. For example, you might say, "You are a careers adviser. Give me three job options for someone who enjoys working outdoors, in a bullet-point list." or "You are an experienced chef. Suggest a healthy dinner recipe for four people using chicken, spinach, and sweet potatoes, formatted with ingredients list first followed by step-by-step instructions." This technique works across most scenarios and the clearer your prompt, the better results you will get.
Get into the habit of asking for help: Whenever you’re stuck on anything - whether it’s a work-related problem, planning a holiday, writing a tricky message, or working out how to fix a dripping tap - try using AI to see if it can help you. You'll be amazed at what it can do, and while it might not solve everything, it can often point you in the right direction or at least give you a helpful first draft.
Don’t take AI responses at face value: Even when AI sounds confident, it is not always correct. Get into the habit of checking important answers, especially if they involve facts, names, or numbers. AI can be incredibly helpful, but it still needs human judgement alongside it.

I've also found following experts on LinkedIn to be a useful way of learning. There are many self-proclaimed experts on LinkedIn and most of them regurgitate information from elsewhere, so it's important to separate out the real experts who produce original and useful content. I highly recommend following both Allie Miller and Ethan Mollick. Allie is the most followed person in AI business with over 2m followers. She presents AI in a very accessible way, and provides lots of practical hands-on tips. She also has a simple email course called AI with Allie which is a great way to get started with AI. Ethan is more academic, and shares lots of interesting uses and research, and is particularly useful to follow for those of you who work in the education industry. His free AI in Education Coursera course is also highly recommended and only takes a few hours to complete.

The Red Giant says

AI literacy is a crucial skill for all of us. Thankfully it's not difficult or technical to learn, and most people have a lot of fun learning how to use the AI tools. Whatever stage you're at, I encourage you to put aside some time each week to experiment with AI, and you'll be amazed at what you can do and where it can help. AI tools are evolving quickly, but the fundamentals of AI literacy will remain true regardless of which AI model you use. The image generation ability of ChatGPT has improved hugely over the past couple of months, so I recommend giving that a try if you haven't already done so. Also, remember that AI doesn't only have to be about work either and it can be used for fun too - one of my favourite AI tools is Suno which gives users the ability to create songs from a single line of text!

Coming Up Next

Next time in The Red Giant I'll be exploring why so many people aren't getting the most out of AI and provide practical tips to enhance your AI productivity. In future editions, I'll also address common concerns about AI including privacy considerations, job displacement fears, and how to evaluate the reliability of AI-generated information.

AI: the most powerful tool in human history

Thu, 24 Apr 2025 09:00:00 +0000

Welcome to The Red Giant

This is my first article for The Red Giant, a newsletter exploring the development of AI and the impact it’s having on education and the future of work. Each article will share my personal perspectives on these topics, alongside an accessible look at what’s happening in the world of AI and how it’s already reshaping how we learn and work.

A bit about me (for anybody who doesn't know me)

I'm Rufus Curnow, an experienced product leader turned AI consultant with over 20 years of experience creating digital learning products. I'm passionate about new technologies, data, and education, and particularly interested in how AI can be used responsibly, working alongside human collaboration and creativity to drive positive change in society.

Why The Red Giant?

Partly, it’s a nod to my name and my height - Rufus means ‘red-haired’, and at 6'8", I’m hard to miss in a crowd. But it’s also a fitting metaphor for this moment in time: Like a red giant star, AI started small but is now expanding rapidly, casting its glow across nearly every part of society. This newsletter aims to explore how this transformative technology is reshaping our world, starting with understanding its unprecedented potential.

AI: The Most Powerful Tool in Human History

As an AI optimist, I firmly believe that recent developments in AI have made AI the most powerful tool in human history. Artificial Intelligence now has the potential to transform society for the better, more so than any technology that came before it.

AI will disrupt each and every one of us, across every industry, and for many people that disruption has already begun as we’ll see below. I’m very aware that there are many challenges and risks that come with AI, but if we can succeed in overcoming those challenges and managing the risks, AI offers us the opportunity to take huge leaps forward in innovation, creativity, and efficiencies. And, just maybe, it could help us build a more equal and prosperous society for all.

This belief in AI’s potential isn’t mine alone and it’s widely shared by those at the forefront of technology:

“The development of AI is as fundamental as the creation of the microprocessor, the personal computer, the Internet, and the mobile phone.” Bill Gates

“Over time, AI will be the biggest technological shift we see in our lifetimes. It's bigger than the shift from desktop computing to mobile, and it may be bigger than the internet itself.” Sundar Pichai

“Success in creating AI could be the biggest event in the history of our civilisation. But it could also be the last, unless we learn how to avoid the risks.” Stephen Hawking

Recent examples of AI impact

There are literally hundreds of examples of where AI is already making a difference to the ways we work and learn, and here are just a few recent examples I wanted to share with you as early examples of AI's potential.

Artificial Intelligence in Higher Education

Faculty are already using AI in many ways to help save themselves time and to save their institutions money. According to the Digital Education Council Global AI Faculty Survey 2025, the most common use case for faculty using AI is in creating teaching materials. Using AI to help create course materials saves faculty time which can be spent on higher value interactions with students, and also enables these course materials to be easily personalise to their specific courses. The graph below shows the top 6 use cases of AI by HE Faculty according to the DEC survey.

It's not just faculty using AI, and a recent study from Kortext found that 92% of students are using AI in some form, with 88% having used generative AI for assessments. Use cases among faculty and students will continue to evolve, and so too will courses and assessments, offering the opportunity for higher education to become more engaging and future-ready, with authentic, skills-based learning that better prepares students for the real world.

Artificial Intelligence in Creative Industries

An AI filmmaker, Pj Accetturo, recently created a The Lord of the Rings trailer in the style of Studio Ghibli. This specific example raises many questions about Intellectual Property rights and creativity, but as a proof of concept for what can be achieved with AI, I think it’s remarkable for just 9 hours work and a cost of $250. The project showcases how AI can democratise creative production, allowing independent creators to produce content that would previously require entire studios and substantial budgets.

See Pj's LinkedIn post for details of how he did it and to view the trailer.

What makes these AI developments in creative industries particularly significant to me is how they're redefining creative workflows rather than replacing human creativity. We will see an abundance of AI generated content, or ‘slop’ as it’s becoming known, but I'm confident that quality will continue to rise to the top. I expect many of the most successful applications will pair human creative vision with AI's ability to rapidly generate and iterate, and, just this week, the Academy of Motion Picture Arts and Sciences has said that films created with AI will be eligible to win Oscars.

Artificial Intelligence in Healthcare

AI is already making a tangible difference in diagnosing diseases, particularly in areas where resources are limited. A recent breakthrough in Benin demonstrated how AI can improve early detection of tuberculosis using lung ultrasound images, a method that’s faster, cheaper, and safer than traditional chest X-rays. Researchers trained an AI system to analyse lung ultrasounds and detect TB with a level of accuracy outperforming human experts by 9%. In a region where access to such specialists is scarce, this represents a major step forward. It means more people can be diagnosed earlier and start treatment sooner, which not only saves lives but also reduces transmission.

There are just a few recent examples of how AI is being used to transform industries, and there’s much more to come as we become more aware of the capabilities of modern AI tools.

The Red Giant says...

AI is an immensely powerful tool which is already making a meaningful impact across many industries, and for businesses of all sizes, it’s increasingly becoming a case of disrupt or be disrupted. We're only just beginning to see the potential of AI, and we need to grow our understanding and awareness of AI before we're able to properly take advantage of the opportunities AI presents. Many people have legitimate concerns about AI - whether it’s the spread of misinformation, environmental impacts, intellectual property and data issues, or the cognitive impact on how we learn to name just a few - and I will explore those in future articles.

However one thing is clear to me: AI is here, and it’s here to stay. Developing our AI literacy and capability won’t just help us manage the risks. It will also put us in a better position to seize the opportunities this technology presents, for individuals, organisations, and society as a whole.

How is AI already changing your work? What opportunities or challenges do you see on the horizon? I'd love to hear your thoughts on AI's impact in your field, so please connect with me on LinkedIn, comment below, or subscribe to this newsletter to join the conversation.

Coming Up Next

In upcoming issues of The Red Giant, I'll be diving deeper into the concerns mentioned above, examining the ethical implications of AI-generated content, analyzing environmental impacts of large language models, and exploring how AI is reshaping cognitive development and learning processes. I'm also intending to share practical frameworks for organisations looking to develop their AI capabilities and implement AI responsibly, and to interview frontline educators and industry leaders about their AI experiences.