Every large organisation I observe right now has an AI transformation program. Most of them will disappoint. Not because the technology underdelivers — the technology is, if anything, ahead of what companies can absorb — but because transformation is being treated as a procurement exercise rather than an operating model change.
The typical pattern looks like this. Leadership announces an AI initiative. A team is formed. Vendors are evaluated. A handful of pilots launch — a chatbot here, a document summariser there. Six months later, the pilots work technically but nothing about how the company operates has changed.
The pilots become permanent pilots. The program quietly loses momentum, and the next budget cycle reallocates the money.I think the root cause is a misunderstanding of what transformation actually means at enterprise level. Buying AI tools is not transformation. Transformation is when the workflow itself is redesigned around the assumption that machines now handle a layer of work that humans used to do.
That requires touching things companies hate touching: role definitions, approval chains, performance metrics, and the informal power structures built on top of them.
Consider a simple example. An AI system can draft a client proposal in minutes. But if the proposal still needs to pass through the same four-person review chain that existed when drafting took two days, you have saved almost nothing. The bottleneck simply moved. Real transformation would ask why four reviewers exist, what risk each one actually mitigates, and whether the AI plus one accountable human achieves the same outcome. That question threatens people. So it rarely gets asked.
There is also a data dimension that gets underweighted. Enterprises sit on enormous information assets, but most of it is fragmented across systems, inconsistently labelled, and governed by nobody in particular. AI exposes this brutally. Models trained or grounded on messy internal data produce messy results, leadership concludes the technology is overhyped, and the deeper lesson — that the data foundation was never built — goes unlearned.
What does the successful version look like? From what I can see, three things distinguish the organisations making genuine progress. First, they pick a small number of workflows that matter commercially and redesign them end to end, rather than sprinkling AI across everything. Second, they assign ownership to operational leaders, not innovation teams — the people who own the process must own its transformation. Third, they invest in the boring layer: data quality, access management, integration. The glamour is in the model. The value is in the plumbing.
There is one more uncomfortable truth. Transformation at this scale produces winners and losers internally, and pretending otherwise breeds quiet resistance. Some roles will shrink. Some skills will lose market value inside the company. Leaders who name this openly, and pair it with serious reskilling, get cooperation. Leaders who hide behind vague slogans about augmentation get sabotage by inertia — the most effective form of sabotage there is.
My forward-looking bet is this: within a few years, the gap between AI leaders and laggards in the same industry will be visible directly in margins, not just in press releases. And when analysts dig into why, the answer will rarely be the model anyone chose. It will be who had the courage to redesign the organisation around it. The question every executive should sit with is simple: are we transforming the company, or just decorating it with AI?
The above reflects my personal views only and is intended for informational and discussion purposes. It does not represent the position of any employer or organisation.