The Talent Was Fine. The System Ate It.

A support call comes in. Somebody on the phone describes, in plain words, a problem worth millions if anyone traced it back far enough. Here is what actually happens to that sentence once it is inside a big company.

The agent logs it. A week later it is a line in a report. A manager reads the report and adds it to a backlog. The backlog gets triaged once a month. The meeting produces an action: investigate. The investigation lands with a team, who write a document. A committee reads the document and asks for more data. Three months have gone. The original sentence is buried under four layers of translation, and whoever actually decides what gets built has never once spoken to the customer who said it.

This is not incompetence. It is architecture. The system was built to manage risk and protect quality in the business you already have, and it does that job well. What it was never built to do, and cannot do by design, is move fast on something new.

I have watched this from the inside. The last three years have gone on exactly this: AI, innovation and adoption in enterprise settings, for a client I am not naming, by agreement. There is plenty to show you. I am simply not allowed to. On one programme, the number of passes it took to get from a stated requirement to a signed-off architecture fell by seventy-five percent, by the client’s own count. That is not a consultant’s slide. That is what the wiring costs when nobody is allowed to touch it.

Enterprise governance exists for good reasons. It stops expensive mistakes, keeps you the right side of the regulator, protects the brand, and holds the quality bar customers expect. For the business you already run, it earns its keep.

For a new venture, it is a death sentence. Every safeguard that protects the core business adds weight to the venture’s ankle. Procurement that takes six weeks. Architecture review that takes a month. A security assessment that outlasts the entire build. A hiring process so slow you cannot get the team in the room for a quarter.

It does not die in one dramatic moment. It dies from a thousand small delays, each one perfectly reasonable on its own, fatal together. By the time the team has fought its way through the machinery, the market has moved on, the window has shut, and the budget review has arrived to finish the job.

Enterprise innovation programmes follow this path often enough that the track is worn smooth. It is almost dull to watch.

A startup hearing the same thing from a customer would move in days. The person who hears the problem and the person who can fix it are sat at the same desk, maybe the same chair on a bad week. No report, no backlog, no committee. Just a conversation, a decision, and something live by Friday.

The startup has no governance, no procurement, no architecture board to answer to. It also has no revenue, no brand, no distribution and no infrastructure worth mentioning. It has exactly one thing: speed. And speed on its own is often enough to beat an enterprise with a hundred times the money behind it.

Startups are not better than enterprises. They are just unencumbered. The real question is whether a big company can build a structure that keeps the enterprise’s resources and steals the startup’s speed. It can. But only if somebody designs it on purpose. Hoping culture will do the job governance will not is how most of these end up.

The case for doing this is financial. Nobody on your board has ever been moved by the word innovation. Carry on running innovation through the normal governance and you already know roughly how that ends, and roughly what it costs. A skunkworks does not guarantee success. What it guarantees is that failure arrives early and cheap, with something learned from it, instead of late and expensive with a deck explaining why.

There is no rail of market statistics here, and that is deliberate. You have your own number for what the last programme cost, sitting in a folder somebody would rather not open. It is more persuasive than anything I could quote at you.