The One-Out-of-Six Problem: Running AI on a Legacy Operating System

Feb 28

Lessons from Davos: Most organisations optimise the technology. Very few re-architect the system around it.

A global retailer built an AI-powered demand forecasting model. The accuracy improvement was massive. The estimated business impact: 4% of company revenues.

They ran it through winter. They ran it through summer. The model performed exactly as designed.

And yet — years later — they are still drowning in inventory problems and working capital issues.

The AI worked. The organization around it didn't.

Nobody changed the process. Nobody set the governance. Nobody built the capabilities that would let people actually trust and use what the AI was telling them.

The board's conclusion? "AI is just hype. The technology isn't there."

Except it was. It was there the whole time.

This story repeats everywhere

This story is not unusual. I hear versions of it almost every week — in manufacturing, insurance, pharma. The details change. The pattern doesn't.

If you've been in a room where a pilot was celebrated and then quietly shelved, you know it. The technology demonstrates value in a controlled setting. Someone presents impressive numbers to the leadership team. And then... nothing scales.

BCG reports that 74% of companies struggle to extract tangible value from AI. McKinsey finds that workflow redesign is the single strongest driver of EBIT impact from AI — yet only 21% of companies have actually redesigned workflows. And a growing number of firms are quietly shelving AI initiatives altogether.

The problem is not technology. It's everything around the technology.

The retailer nailed their data and their technology. But process redesign? Untouched. Governance? Nonexistent. People and culture? Nobody asked the frontline teams whether they'd actually change how they work based on AI outputs. Leadership alignment? The board never owned it. Strategic clarity on why this mattered? Assumed, never articulated.

AI readiness has six structural dimensions. The retailer solved one.

One out of six. That's why it failed.

Six questions most leadership teams avoid

There are six questions that I use to predict whether an AI initiative will scale or stall. Ask leadership teams about their tools and pilots — most have an answer. Ask about adoption and AI readiness, and the room goes quiet.

Most have activity. But very little momentum.

"What does your organization need to become — and how does AI get you there?" Not which tools to introduce. Not which cost center this will be allocated to. Where is the business going, and what role does AI play in that future? Without an answer, every pilot is an orphan fighting for budget against whatever feels urgent this quarter.
"If I asked your middle managers whether AI has made their work better — what would they say?" A digital transformation leader I spoke to recently shared a telling exchange: "I asked a senior manager — are you busier now than before? Yes. Is the quality of what you're reading better or worse? Worse." AI makes it easy to produce more. So people produce more. But nobody's asking whether any of it is good. Here is the issue: 93% of AI budgets go to technology — while 70% of the value comes from people and processes.
"Which workflows have you actually redesigned — not just automated?" This is where the retailer's story lives. The planning cycle, the approval chain, the handoffs between teams — if none of that changes, the AI just sits there being accurate while the organization keeps doing things the old way.
"Who is accountable when AI gets it wrong?" Nearly half of employees upload sensitive data to public AI tools. Two-thirds don't verify outputs. Most organizations have no policy. The EU AI Act is adding pressure from the outside — but the real issue is internal. Without trust infrastructure, people simply won't act on what AI tells them. This is what we call Rebuild Governance — the third step of our REWIRE methodology.
"Can your data actually support what you're trying to build?" The data usually exists. It's in silos, inconsistent formats, owned by nobody. 60% of AI projects will fail without AI-ready data. That's not a technical fix. It's an organizational one.
"Is your CEO using AI — or just sponsoring it?" Someone put it bluntly: "There's not one board member that really understands AI. Not one. They pretend because they don't want to look stupid." Leadership doesn't mean signing off on a budget. It means using AI in your own work and demanding the change from your teams. Being role models. Not giving mandates. Not sending out memos. Role models. Our executive bootcamp s are designed for exactly this.

What intentional looks like

Now contrast the retailer with a large American pharmaceutical company.

Their CEO named AI as one of four company priorities for 2026 and beyond. AI goals cascade into every senior leader's performance review. Leaders are evaluated — formally — on how well they advance AI adoption in their function. There need to be incentives for individuals to use AI.

Not subtle. Not optional. And it works because AI isn't treated as a technology project — it's an organizational capability. It works because it touches multiple dimensions at once: leadership owns it, people are incentivized, processes are being redesigned around it, and governance is built into how the organization already operates. Strategy, people, process, governance, data, and leadership all moving at the same time.

That's not pilot purgatory. That's intentional transformation.

So — where do you actually stand?

Most companies can describe their tech stack in detail. Which models they're testing, which vendor they're evaluating, which use case they piloted last quarter.

Very few can answer all six questions honestly.

That's the real gap. Not between companies that use AI and companies that don't — but between those that have addressed one or two of these and those working on all six.

The retailer had world-class AI. They had the data. They had executive interest. But without the process change, the governance, the cultural readiness, and the strategic clarity, the AI just sat there being accurate while the organization continued operating the way it always had.

Take the AI Readiness Check — click the banner above.
AI maturity is not about how advanced your tools are. It’s about whether your organisation runs differently because of them.

Most companies upgrade the technology. Very few upgrade the operating system.

If you don’t know which dimension is holding you back, you can run a check here: