Closing the Gap Between AI Investment and AI Impact with Max Haining, 100 School

John Emmerson: Today I’m delighted to be joined by Max Haining, founder of 100 School. For people who are unfamiliar with 100 School, why don’t you tell us a little bit about what you do?

Max Haining: 100 School is an AI capability organisation that goes into companies and helps them raise the baseline of their AI capability. Our main ethos is helping people change their behaviour and, as a result, helping teams change how they work with AI — in daily bite-sized chunks, which is our learning methodology.

John Emmerson: There are so many organisations trying to do something with AI without actually increasing their capabilities and failing. There’s a stat from the AI Enablement Insider survey showing 72 per cent of organisations that have had capability training are seeing measurable ROI, and only two per cent of those that haven’t had training are seeing any. It’s clearly important.

Max Haining: Yes, and what’s interesting is we’re seeing different phases of AI capability. The last two years have been organisations, understandably, chucking mud at the wall. They’ve got this big buzzword, all this noise, and they need to do something about it. But there hasn’t necessarily been a best-practice approach because we’re all trying to do this for the first time. There’s been no manual playbook to follow, hence why a lot of mud has been thrown — some of which probably shouldn’t have been. But in the act of trying to move the needle, or trying to look like you’re doing something with this generational technology, there’s been a lot of mess to pick up.

Max Haining: So we’ve seen two years of experimentation, and now lots of companies are thinking: we’ve done this experimentation, a lot of mud is at the wall, it hasn’t necessarily moved the needle. How do we take this big investment — probably some of the biggest investment we’ve put into software — and turn it into results? We’re working with a few companies right now and they’re sharing with us: this is the year of AI impact. That is easier said than done, but generally we’re entering the phase of: we need to make our money’s worth from actually applying this technology and seeing the results.

John Emmerson: How do you stay on top of it, given how fast the pace of change is?

Max Haining: The default approach that many companies take is: we’ve got all these tools coming out, all these new features being released every week, so let’s just keep adding new workshops to try and keep up. That’s the MO we’ve seen. The problem with that is people are overwhelmed trying to process AI in its current state, let alone the latest features being released week by week. It adds more chaos and anxiety to your team. And it’s not sustainable — it feels like keeping up, but it isn’t.

Max Haining: What we like to do is go two steps back from the features and the tools and ask: what behaviours do you need to learn in order to be good with AI, so that even when a new feature releases, you’ve got that transferable skill set that allows you to learn it yourself quickly? You’re agile enough to learn it versus waiting for the next hit of a specific tool tutorial. If you have to wait for that — if that’s spoon-feeding — your organisation isn’t agile enough to keep up with the pace of change. Build the behaviours. The features will follow.

John Emmerson: I understand your team went on a pilgrimage to San Francisco last year to meet companies including Anthropic and Meta. What came out of that?

Max Haining: Yes. We went to the ASU GSV conference in San Diego first — one of the world’s largest education conferences where everything from Harvard down to new EdTech start-ups comes together. The seed was planted there. Everyone was talking about AI and helping people adopt it. But what we found interesting was not what they were saying — it was what they weren’t saying. Nobody was asking: what does good actually look like? Before we even start learning this, before we start teaching it, what is the benchmark for what it means to be genuinely great at AI as a person and as an organisation? There wasn’t a shared baseline.

Max Haining: That got us thinking: before we dare to start teaching teams on this technology, we need to understand what separates those who are using it at the frontline exceptionally well from those who aren’t. What are the behaviours, skills and mindsets that can help anyone close that gap? So we went to San Francisco on what I’d call an observational tour — different companies, hackathons, cracked founders who are in the top one per cent of users. We took all those insights and behaviours and thinking patterns and built our own framework, what we now call the AIQ framework. That is the backbone for any curriculum or educational support we give to companies and individuals.

John Emmerson: Tell me more about the AIQ framework.

Max Haining: The framework is built on what are the most durable, tool-agnostic skills that can apply anywhere, to anything, to any problem — the skills someone needs to extract the real benefit of AI. It is made up of 12 behaviours across three core ingredients. The first is recognising an opportunity when you see it. A lot of people struggle with imagining what they can even use AI for in the first place — I know that’s something your research at aibl has reflected as well. The second is communication: if you’ve got a great use case, how do you actually communicate with AI to get a good output and good results from it? Communication is a big piece of this. The third is evaluation: how do you evaluate what AI gives back? How do you own the output?

Max Haining: I saw someone recently talking about the trend of AI dumping — where people just pass on a bit of work that Claude or another tool has produced, as if to say: here’s something the AI spun up. Is that your thinking or is that someone else’s? Those three components — recognising, communicating, evaluating — are what we bake into every lesson and every touchpoint with what we teach.

John Emmerson: The AI dumping issue is real. It’s important for people to make sure they’re not just passing off AI outputs as their own, and to actually check and iterate on what comes back. You mentioned the Anthropic AI Fluency Report in relation to this.

Max Haining: Yes. The Anthropic AI Fluency Report showed that people using Claude to create artefacts have a high propensity to believe the output — especially when it’s presented as an artefact, because it looks polished and visually well-presented. You’re more likely to check a plain text output. You’re less likely if it’s been visualised or looks nicer. That is a dangerous pattern of behaviour that is very easy to fall into, and that I myself have been guilty of. Being aware of those patterns — regardless of whether you want to fall into them or not — is really important as part of the check and balance.

John Emmerson: What does it mean to be AI native today, especially for an established team trying to rewire how they work?

Max Haining: I think the definition is ongoing and will evolve as the tools and capability change. But at the core of it, I see it as a set of habits — an order of habits where your team is habitually using AI in a way that they’re proactively and systematically applying it to their individual tasks. And then collectively using skills or versions of that — whether in Claude or other large language models — that centralise shared context and data to continuously improve not only the speed but the output of their work.

Max Haining: Ultimately, the best teams I’ve observed using AI — I call them the Navy SEALs of AI — feel like they’re living in the future. The fundamental difference is they are not doing any of the same tasks they were doing two years ago. They have abstracted themselves into a new layer of work and let AI do the grunt work, the grisly stuff. They’re now allocating their time completely differently. That is what AI native actually looks like.

John Emmerson: Is your programme primarily a learning product or a behaviour change product?

Max Haining: We frame it as a behaviour change product that happens to deliver content. Behaviour change is the number one goal. Our anti-goal is to create more content. Our anti-goal is to be another B2B learning provider where we create custom content that lives and dies in a matter of weeks based on the state of change externally, that no one uses and that no one actually changes their behaviour as a result of.

Max Haining: What we’re trying to do is help people unlearn certain behaviours they’ve built up for years — pre-AI — to then reimagine their work and their role in an AI age, in and amongst all the chaos of their existing job and existing targets. To try and drive that change in all that chaos is genuinely hard. So the way we accommodate for that context is to come in and run a 15 Days of AI programme with a team, which is normally a company-wide initiative. There’s a big communications run-up so everyone’s aware, ready and — most importantly — excited. Most AI training is mandated. We make a point of sharing with our clients: this needs to be voluntary, it needs to come intrinsically from your people.

Max Haining: Our most successful programmes have been run over seven weeks, with a cadence of two lessons per week dropping into people’s inboxes. Each lesson takes around 15 minutes — a little snack in a coffee break. In each lesson, they learn a concept and a framework, and then apply it right there and then to their work that day. So it instantly hits them with value, rather than messing about with something separate from their flow. That’s also what keeps people coming back to the next lesson in the series. And then the last layer — which I think is probably the most important part — is the cultural layer. Side by side with all of this, we run a Slack or Teams channel where everyone shares what they’re learning. That becomes visible to the whole organisation. If people aren’t on the train, they see this groundswell of energy, people learning and having those wow moments, and start to reconsider whether they should be engaging with AI. That’s where you can pull the rest of the organisation into a more AI-native mindset.

John Emmerson: Why 15 minutes specifically?

Max Haining: Fifteen minutes is short enough to be attainable and realistic, and long enough to extract a useful outcome for your working day. With AI, because it’s so fast, that is genuinely achievable. Pre-AI, it was harder to do that in that space of time. And ultimately, people are busy, their jobs are high-intensity, so 15 minutes becomes manageable enough in a coffee break or lunchtime for most people. But it’s still an ask. It’s still an ask.

John Emmerson: When people hit a wall they can’t get past — when the tool doesn’t integrate with a system or something breaks — how do you help them?

Max Haining: There are a few pieces to this. At a very practical level, we always encourage people to ask AI first. Just to help them through the problem — and to build that AI-native spirit, you have to default to asking AI. Beyond that, once you move past the individual layer of productivity, you’re in the team layer, which collides with teammates, processes, tool access and all the rest of it. Some of that is outside your control. So the question becomes: how do we help people bring their team along to this? Some things you want to build are best built as a prototype first, which you then hand off to your product team or developers. Your job in some cases is not to build the full thing — it’s to build the imagination so that others on your team go: that is a great idea, let’s build it. And many organisations are running what I’d call champions programmes — the evangelists inside the organisation, the technically most proficient people, forming that SWAT team for AI culture. Those are the people you’d want to tap if you hit a wall.

John Emmerson: Final question. You’re sitting across from a sceptical mid-market leader — limited resources, they’ve seen learning programmes come and go with no measured impact. What’s your closing argument?

Max Haining: You have already invested a lot of money. You’ve given people access to tools. We’ve moved past the acknowledgement problem — you’ve already accepted this technology is good enough to give to your team. The problem is you haven’t given them an activation point that helps them structurally implement this into their everyday work. There’s a big gap between the investment you’ve already spent and the impact you’re seeing.

Max Haining: To start closing that gap, you need to do two things. First: give people a structured way of learning this that is digestible and embeddable in their daily flow. Give them pathways. You can’t give them chaos. You can’t give them workshops scattered here, there and everywhere. Give them structure, pathways, progression, a journey. Second: create a measurement arm in your organisation. We can talk about skills all day, but how do you know they’re moving the needle? How do you know your people are reaching a shared level of AI fluency? Do those two things. That is when the scepticism starts to shift.

Why do most AI training programmes fail to produce measurable ROI?

Max Haining’s diagnosis is that most organisations have given their people access to AI tools but have not given them a structured activation point — a way to systematically implement that access into their daily working patterns. The result is a large gap between investment and impact. Two years of experimentation, workshops and tool rollouts has produced what he calls mud at the wall: a lot of activity, limited evidence of changed behaviour. His argument is that access without structure produces nothing measurable, and that the organisations now demanding ROI from their AI spend are discovering this the hard way.

What is the AIQ framework and how was it developed?

The AIQ framework is 100 School’s 12-behaviour model for AI capability, built from an observational research trip to San Francisco studying the top one per cent of AI users — founders at hackathons, teams at leading AI companies, people who are genuinely at the frontier of day-to-day AI use. The framework deliberately avoids being tool-specific or feature-led. Instead, it identifies three core components that apply regardless of which tools or models someone is using: the ability to recognise an opportunity for AI, the ability to communicate with AI effectively to get a useful output, and the ability to evaluate that output critically rather than accepting it at face value.

What does it mean to be genuinely AI native?

Max Haining’s definition is concrete and deliberately demanding: a team is AI native when its members are not doing any of the same tasks they were doing two years ago. Not faster. Not better. Different. The teams he calls the Navy SEALs of AI have abstracted themselves into a new layer of work — they have let AI do the grunt work and are now allocating their time to work that requires judgment, creativity and relationships. This is a much higher bar than the common usage of AI native, which tends to mean anyone who uses ChatGPT regularly. Max’s version requires not just habit change but role redesign.

What is AI dumping and why is it a risk for organisations?

AI dumping is the practice of passing on AI-generated output — a draft, a summary, a document — as a contribution without having critically evaluated, owned or meaningfully shaped it. Max Haining references the Anthropic AI Fluency Report, which found that people are significantly more likely to accept AI output without checking it when it is presented visually as a formatted artefact. The polished appearance creates a false sense of authority. The risk for organisations is that AI dumping erodes quality standards quietly — outputs look complete and professional but the thinking behind them is shallow or absent. Building evaluation as a habit, not an afterthought, is the counter.

Why does 100 School use 15-minute learning modules rather than intensive workshops?

Fifteen minutes is short enough to be realistic for people with demanding day jobs, and long enough to learn a concept, understand a framework and apply it immediately to real work that day. The immediate application is important: it means the lesson delivers visible value in the working day it is taken, rather than something to be applied later in theory. Max Haining also notes that AI itself makes 15-minute sessions more achievable than pre-AI learning, because the speed of iteration means you can actually practise and see results within that window. Intensive workshops, by contrast, are consumed once and rarely change sustained behaviour.

What is the cultural layer of AI capability building and why does it matter?

Alongside the structured learning content, 100 School runs a shared Slack or Teams channel where programme participants post what they are learning and experimenting with. This shared visibility serves a specific function: it creates social proof and peer pressure that pulls less-engaged colleagues towards an AI-native mindset. People who were not initially on board see the energy and the momentum building, and reconsider their own engagement. Max Haining argues this cultural layer is probably the most important part of the programme — more important than the content itself — because behaviour change requires visible cultural pressure from peers, not just individual learning.

What two things does a mid-market leader need to close the gap between AI investment and AI impact?

Max Haining’s closing argument to sceptical leaders is direct: structure and measurement. Structure means giving people pathways — a coherent journey of digestible, embeddable learning with clear progression — rather than scattered workshops and tool access with no connective tissue. Measurement means creating a way of knowing whether people are actually reaching a shared level of AI fluency, and whether that fluency is changing how they work. Without structure, people remain stuck in the same patterns despite having access to the tools. Without measurement, the investment stays on the cost side of the ledger indefinitely because there is no way to demonstrate that anything has changed.