Tools nobody owns
Each one understood by the person who built it. Nobody wrote down what it is for. When they leave, it goes quiet.
McKinsey's global survey of 1,719 respondents, published August 2026. The first answer is the one everybody reports. The second is the one a board asks about.
"We have not gotten enterprise-level productivity, with the exception of a few areas."
Eric Kutcher, chair of McKinsey North America, October 2026They built what they could see, with no system to build into. Four things follow, and all four are structural.
Each one understood by the person who built it. Nobody wrote down what it is for. When they leave, it goes quiet.
Written in an afternoon, reviewed by no one, tested in no structured way. You are trusting, not knowing.
Spend is visible, yield is not. Nothing connects the cost to the job it did.
Every build starts from nothing. The tenth tool costs what the first one did.
What goes in, what it does, what you get back — and the loop that makes the next one cheaper than the last.
What your AI spend actually bought, job by job. Bills tend to fall once someone can see them, and we learn how the company really works before we build anything.
The business understood properly, one problem taken all the way to live, proved and handed over — with your people trained to do the next one.
Most clients ask us to keep operating it. Some take it in-house entirely. Either way it belongs to you, and what we charge is tied to results.
The nine stages, who does what at each, and every check a change has to pass.
One thing that costs more time or money than it should. We'll tell you whether AI is the right answer — and we'll tell you when it isn't.
Thirty minutes on one problem is usually enough to tell whether we can help. If we can't, we'll say so.