Every board has discussed AI. Most have approved budgets. Many have run pilots. And yet, across virtually every sector we examine, the number of companies reporting material improvement in their P&L as a result of AI investment remains stuck in the low single digits.
BCG research puts the figure at 5%. Our own analysis of mid-market companies across the UAE, Europe, and the United States is consistent with that number. A very small number of companies are genuinely using AI to improve profitability. The rest have a more complicated story to tell.
The five failure modes
We have identified five recurring patterns in companies that have invested in AI without seeing it in their financials.
Piecemeal deployment
AI is deployed in isolated experiments rather than across a business function end-to-end. Individual tools show promise in demos. Integration into the workflow never happens. The insight exists. The process doesn't change.
Underfunded to completion
The initial budget covers the pilot. It doesn't cover the integration, the data work, the change management, or the iteration required to reach production quality. The project stalls at 60% complete and is declared a success anyway.
No value-first blueprint
AI is deployed without a clear definition of the financial outcome it is supposed to produce. Usage metrics are tracked. P&L metrics are not. Without a direct line between the AI deployment and a specific financial result, the connection is never established.
Tracking motion, not results
Success is measured in AI outputs — documents processed, queries answered, content generated. Not in the downstream financial impact of those outputs. The AI is productive. The margin doesn't move.
People and organisation ignored
The technology is deployed. The workflows are not redesigned. The team continues to work around the AI rather than with it. Within six months, usage falls and the implementation is quietly shelved.
What the 5% do differently
The companies seeing material P&L improvement from AI share a set of practices that distinguish their approach from the 95%. They are not secrets — they are disciplines.
They define the financial outcome first and work backwards to the AI deployment, rather than deploying AI and hoping a financial outcome emerges. They fund transformation to completion, not to pilot. They redesign the workflow, not just the tooling. They measure success in margin improvement, not in AI activity metrics. And they maintain senior attention on the programme through to the point at which the financial result is locked in — not just through launch.
The pattern is consistent across sector, geography, and company size. The companies extracting financial value from AI are doing something structurally different from those that aren't. The gap is not technical. It is organisational and financial.
The investment implication
For investors, this creates a clear thesis. Companies that have figured out how to make AI pay are being repriced. Companies that haven't are still priced on their current earnings. The gap between those two valuations is significant and widening.
The opportunity is not in buying the companies that have already made the transition — that premium is priced. The opportunity is in finding the companies that will make the transition, and being the partner that makes it happen.