The Transformation Gap: Why Most AI Programmes Don't Land, and how to actually succeed
- Aug 14
- 5 min read
Updated: Aug 19
The honest state of play today
The adoption of AI is seen as a business imperative. One in six people globally now uses generative AI. Around 85% of organisations say they expect to deploy customised AI agents soon. Most boards have an AI slide, most peers claim an AI strategy, and somewhere in your inbox there is inevitably a pitch for an enterprise licence, a new model, an agent platform, a data layer and a consultant, all promising the same thing.
And yet … the reality is that whilst there is high adoption, there is still low meaningfully business transformation. A PwC 2026 CEO survey found only around one in eight CEOs say AI has genuinely increased revenue and reduced costs. Only 1% of executives describe their company as mature in AI deployment. Most organisations are stuck in pilot mode, or worse, in the comfortable middle where nothing compounds.
The questions we, Grace Blue and Mindstone hear are not the ‘why’ or the ‘what’ but the ‘how’.
The Productivity Gap study in 2026 found that just 20% of organisations capture 74% of all AI's economic gains. The drivers and the dabblers are separating, and the gap is widening, because learning compounds. In the same spirit, such is the pace of change happening around AI, every month of delay is benefiting a competitor.
Why do most AI transformation programmes stall
When AI transformation programmes stall, the same three reasons appear time and again. It’s not infrastructure, technology, or the ones in the press releases….
Mindstone has run AI programmes with some of the biggest organisations in the world, from Home Depot to Pearson to Delta Air Lines. More than fifteen thousand people have been through their programmes, and the average person who completes competency gets back a minimum of five and a half hours a week (it’s a lot higher for agentic). Mindstone's own engineering team runs at fifty times its previous output. The belief is that this is a fraction of what is possible for every single team (often called the Everything Factory). This extensive experience working with a huge array of businesses provides a unique insight into what are the biggest barriers to success.
1. The leadership journey hasn't been taken
Most leaders genuinely believe AI matters yet relatively few have a real understanding of what it’s potential is. That gap is not technical. It is human.
The biggest blockers in any organisation are people dynamics dressed up as business cases. Concern, worry, fear of getting it wrong, embarrassment at admitting you don't know. They arrive at the table as perfectly logical, risk-based decisions. And the more senior the leadership, the more dynamics you are navigating. If the person signing the AI budget isn’t using AI to its full potential, they risk making the wrong (and expensive) decisions.
2. Delegation leaves it to culture
"I don't have time, so I've hired a Head of AI (or given it to a team)." It sounds sensible. It is the single most common failure mode we see.
When you delegate transformation, you are outsourcing the one thing that cannot be outsourced: the navigation of people through change. You are asking a person or a team to navigate something they have never navigated before, which means failure after failure. When you delegated it, you are at the mercy of your culture, which was the thing you were trying to change.
3. Upskilling is being mistaken for transformation
“Appoint a core team to solve X, Y, Z, plus a bit of training rolled out for the masses”. It is the most common pattern in the market, and it is the reason most companies are not getting the results they could.
Skills are about 10% of transformation. Behaviour is about 90%. Tools sit unused not because people can't be trained, but because nobody changed how work actually happens, and nobody gave permission to change it. If you control too tightly, too quickly, employees quietly use consumer tools anyway, and you learn about it a year later in a security report.
The ‘how’ that differentiates the winners
Mindstone compressed everything it has learned into three takeaways. If you do these, the rest follows.
1. Lead from the front
It starts with the CEO, and the whole C-suite goes on the journey together. Not a briefing. A journey: understand what is possible, use the tools properly every day, and navigate your own personal 30-60-90 of change, with everything that comes with it. Leaders who lean into behaviour change around an AI transformation programme, adopt and embrace it, are the ones who set the tone for how the rest of the culture behaves.
Taking each of them through that journey is not just upskilling. It is coaching and facilitation. Every leader carries a private version of the same question: what happens to me if I get this wrong? That question rarely gets asked out loud. It arrives as a perfectly reasonable objection, a data security concern, a governance risk, a "let's wait and see" “I think we can do this ourselves”. Some of those are real. Most are easier to say than the truth, which is that they are worried, or embarrassed, or afraid of looking like they do not know.
The job of the journey is to separate the two, honestly. A genuine data security concern gets solved. A fear dressed up as one gets named, and then it stops blocking the room. Only once you have done that can you roll out across the business, because only then are you equipped to lead your own teams through their journey. Your people do what you do, not what you say. If the exec team does not use AI daily, nothing else in this piece matters.
2. Engineer behaviour change, not just skills
Give freedom and experimentation before you impose control. Permission to fail is more powerful than governance, and it is the difference between adoption and theatre. Yes, you need governance. Yes, as you go agentic you need to watch token cost. But these are manageable, and once the senior team has been through a proper process, they are equipped to manage them themselves, without a squad to save them.
Organisations that enable, encourage, motivate, and practice what they preach at every levels are the ones who ultimately benefit from the compound effect of learning.
3. Measure where people are, so you can support
This is the most underestimated piece, and the biggest gap we solve. Most businesses do not measure the right things, so they do not know where they are, and cannot intervene. Where is each team? Each individual? Who is stuck at search-replacement level, who is building agents, who has quietly given up?
The companies that win track the journey, not just the training completion. They see the curve, they spot when it stalls, and they put support exactly where the stall is. Proper measurement and visibility is what turns a one-off programme into a compounding capability, and it is what ‘champion squads’ and ‘cheap training’ never give you.
We are at the start of this, not the end. Every number in this piece will look ancient in three years. The companies that pull away will be the ones where the leadership treated AI as a people transformation, that they personally led, and measured relentlessly along the way.
See the event video below:
Comments