The Ferrari engine in the horse cart

There is a standard recipe for enterprise AI adoption.

Your organisation is probably following it right now.

Take an existing process. Find the step where a human does something repetitive. Replace that step with an LLM.

Congratulations. You've bolted a Ferrari engine onto a horse cart.

The recipe preserves every assumption of the legacy process, from the data model to the human role, and delivers a marginally cheaper version of the same capability. The cart doesn't go faster. The wheels weren't built for the torque. The engine's very real power is spent proving, expensively, that power was never the constraint.

That's not transformation. That's an expensive way to stand still.

In one pilot, an AI agent cut commercial loan approval from 16 days to 8. Then production arrived. Clients still took days to answer follow-up questions. Credit decisions still waited for committees. Approvals still queued. The calendar barely moved; the freed hours went on coffee and social media.

The step got faster. The process didn't.

The alternative starts from a different question. Not "where can we insert AI into what we have?" but: if we were building this process today, knowing what AI can and cannot do, what would it look like?

Those two questions produce structurally different enterprises. The first fills MIT's statistics. The second is the final chapter of my book, and the subject of the article I published this week: why the failure of the first approach is designed in, not bad luck.

The full article is here The Ferrari Engine in the Horse Cart: Why 95% of Enterprise AI Pilots Were Designed to Fail

Now the diagnostic. Look at your current AI initiatives and count the ones that began with an existing process and a step to swap out.

If the answer is “all of them”, that's not your team's failure, it's the industry default. How many of yours began that way?