Why Your AI Transformation Is Failing, and What to Do Instead

This week's Conversation is an essay, not an interview. The Sensing Report documented what 1,200 CEOs told WRITER anonymously about their AI transformation programmes. I want to tell you why they gave that answer, and what to do about it. I am writing this from thirty years inside the machine that produced the strategy they are now calling theatre.

I spent most of my career in Human Capital consulting. I was a partner. I have sat in more CEO offices in Mumbai, Riyadh, Singapore, Beijing, and Dubai than I can count, presenting workforce transformation roadmaps with measured confidence. Some of what we sold was right. A great deal of it was sold because we had sold it before.

So when the WRITER 2026 survey, released this month, reports that 75 percent of C-suite executives now say their AI strategy is "more for show" than actual guidance, I do not read that as a revelation. I read it as an overdue admission. Having spent three decades inside the industry that built the decks they are now describing as theatre, let me tell you what I think is actually going wrong, and what the small number of organisations doing it right are doing differently.

The four mistakes

The first mistake is treating AI as a procurement decision.

In the 2000s and 2010s, every major transformation programme began with a vendor selection. HR transformation meant picking Workday or SuccessFactors. CRM transformation meant Salesforce or Microsoft. ERP transformation meant SAP or Oracle. The CHRO and CIO ran a process, picked a partner, signed a statement of work, and the implementation began. That logic is now being applied to AI. It does not work.

AI is not a system you procure. It is a capability you redesign operations around. If your organisation's AI programme is being run out of your procurement function or your enterprise architecture function, with a vendor-selected platform at the centre, you have already chosen a strategy that the evidence now says will disappoint. The WRITER number is not random. Of the 1,200 executives surveyed, 59 percent are investing over a million dollars annually in AI technology. Only 29 percent report significant ROI. The gap is not in the spending. It is in the operating model around the spending.

The second mistake is treating AI transformation as a communications exercise.

I have watched this pattern in three regions. The CEO announces an AI strategy at an offsite. The communications team produces a manifesto. A Chief AI Officer is appointed. An AI Council is formed. Training videos are rolled out. A town hall is held. The dashboards begin tracking adoption metrics. Six months later, individual productivity has risen modestly in the marketing and finance functions. Nothing else has changed, because nothing else was designed to change. When the WRITER survey says 75 percent of strategies are "more for show," this is what they are describing. The strategy was the show. The show was the strategy. There was no plan B because there was no plan A beyond the announcement.

The third mistake is mistaking AI proficiency for AI transformation.

The Writer survey has a striking finding. Ninety-two percent of executives say they are actively cultivating "AI elite" employees. Sixty percent plan layoffs for AI non-adopters. The implicit theory is that if you can move enough individuals up the AI-skill curve and remove the ones who will not move, the organisation will transform. This is a category error. Individual AI proficiency produces individual productivity. Organisational transformation requires redesigned workflows, redesigned decision rights, redesigned accountability, and redesigned performance management. The gap between a workforce of AI-skilled individuals and an AI-transformed organisation is where the WRITER survey respondents are stuck. A thousand prompt engineers cannot redesign a P&L structure. Only the executive team can, and most executive teams have not.

The fourth mistake is running AI transformation as a top-down cascade.

The 2026 Human Capital Trends research from one of the large global consultancies, which I have followed closely for most of my career, finds that only 6 percent of organisations report significant progress on designing work for human-AI collaboration. Sixty-five percent say their culture needs to change. These numbers are consistent with WRITER's findings and they tell the same story from a slightly different angle. The organisations failing at AI transformation are the ones running the standard cascade: board decides, CEO announces, C-suite aligns, line leaders execute. That cascade works for cost-reduction programmes and for system implementations. It does not work for AI transformation, because AI transformation requires continuous redesign at the work level, and the people who know how the work actually happens are not in the cascade.

What the minority is doing differently

The minority of organisations that appear to be making AI transformation actually work share a small set of practices, most of which look unfamiliar to a traditional transformation office. I will describe five I have observed in client work across India, the GCC, and Africa over the past eighteen months.

They start at the workflow, not the platform. A Mumbai-based mid-cap financial services firm I know began its AI programme by asking operations leaders to select five workflows where a 30 percent cycle-time reduction would change the P&L. They did not select a vendor first. They did not appoint a Chief AI Officer. They did not run a town hall. They ran five workflow redesigns with embedded AI tools, measured outcomes, and scaled what worked. Eighteen months in, they have redesigned 43 workflows. Their AI infrastructure bill is a fraction of their peers. Their operating leverage is visible in the quarterly numbers.

They fund the transition from the savings, not from the workforce. This sounds obvious. It is vanishingly rare. The Oracle pattern that reAImagine.work covered in Issue 002, and that has since become the industry default, is to fund AI infrastructure by reducing workforce costs. The minority pattern is to redesign first, measure the savings, and fund the next phase of AI investment from those savings without requiring workforce reductions to finance the cycle. This is slower. It is more politically difficult. It produces organisations with durable capability rather than transitional cost bases.

They put the workforce transition inside the operating plan, not beside it. Most organisations still treat reskilling as an HR programme, running alongside the AI transformation rather than being the same plan. The minority embeds workforce redesign into the operating plan, with the same reporting cadence, the same finance discipline, and the same accountability. Hindustan Unilever has done this. Saudi Aramco has done this at the level of its nationalisation plan. The major human capital research houses have been calling for this for two years, and almost none of their clients actually do it.

They accept that AI transformation means restructuring, not augmentation. The comfortable slogan of the past three years has been "AI will augment, not replace." That slogan is now colliding with the WRITER survey reality. Organisations are doing mass layoffs in the name of AI transformation. The minority of organisations making transformation work have stopped using the augmentation language externally, because it is not honest, and started describing restructuring openly. The quality of their internal conversation has risen as a result. The quality of their workforce planning has risen. The quality of their investor disclosures has risen.

They build the measurement system before the programme, not after it. The reason 75 percent of CEOs describe their AI strategy as "more for show" is that they have no measurement system that would reveal it to be otherwise. Individual productivity dashboards, training-completion rates, and AI tool adoption metrics are not measurement systems. They are activity reports. The minority of organisations have built measurement systems that track cycle time, quality, cost per outcome, and workforce composition change quarter over quarter. They know what is working and what is not. They also know their AI transformation is genuinely redesigning the organisation, because the numbers say so.

The framework

Here is what "doing it right" looks like, captured in a framework I have been using with clients this year. It has four layers, and the mistake most organisations make is starting at the top and stopping at the second layer.

Layer one is the workflow. What work is being done, by whom, with what inputs, producing what outputs, at what cycle time and cost. This layer is invisible in most transformation programmes, which is why most transformation programmes fail.

Layer two is the operating model. Decision rights, accountability, performance management, span of control, career paths. AI transformation changes all five of these, and if layer two is not redesigned, the layer-one improvements do not scale.

Layer three is the capability. What the workforce can do, individually and collectively, augmented by AI. This is where most organisations concentrate their effort, which is why most organisations produce AI-proficient individuals inside unchanged operating models.

Layer four is the strategy. What the organisation is competing on, how it is monetising its capabilities, how it is allocating capital. Most AI strategies operate only at this layer, which is why they look like theatre to the workforce and like theatre to the board when asked anonymously.

The organisations that are making AI transformation work are redesigning all four layers in sequence, with the measurement system in place before the redesign begins. They are a minority. They will be the organisations that emerge from this decade with a durable advantage.

One last thing

I want to be honest about why I am writing this essay. I spent thirty years selling the transformation approach that the WRITER survey has now documented as failing. Some of what I sold genuinely helped clients. Some of it was sold because the client expected it, the partner needed the revenue, and the industry had produced no better alternative. I did not have the vocabulary in 2012 to describe what I see clearly now. I am not writing to absolve myself. I am writing because a lot of my former colleagues, and a lot of the clients I worked with for two decades, are now inside the 75 percent number. They know the strategy is theatre. They are looking for something else to do, and they have not yet found someone willing to tell them the truth about why the old playbook is not working.

This magazine is my attempt to tell that truth, weekly, with specific signals from specific companies in specific geographies. The organisations doing AI transformation right are small in number but they are identifiable. Over the coming issues of reAImagine.work, I will name them, describe what they are doing, and explain why it works. If you are running a transformation programme and you suspect you are in the 75 percent, write to me. I read every response.

The admission has been made. The work begins now.