Skip to content

The Contact Test

The Prussian field marshal Helmuth von Moltke observed that no plan survives first contact with the enemy. AI projects have their own version: no model survives first contact with the organization. The technology rarely decides the outcome. The organization does.

The test is not a theory. It is a pattern, and it is well documented. MIT's State of AI in Business study found that 95 percent of corporate GenAI pilots deliver no measurable return — and located the cause in how organizations adopt tools, not in the models. S&P Global counted 42 percent of companies abandoning most of their AI initiatives in 2025, up from 17 percent a year earlier. And the Boston Consulting Group puts numbers on the split itself: successful AI transformations are roughly 10 percent algorithms, 20 percent technology, and 70 percent people and processes. My own version of the same pattern comes from leading data and AI projects at a large Swiss health insurer and from digitalizing SMEs where the budget was the owner's own money. Projects rarely die of bad models. They die of a missing decision, an absent owner, unconvinced users, governance discovered too late, and habits nobody planned to change.

Those are the five forces: Decision, Owner, Users, Trust, Habits. Five questions, answered honestly, before the budget is spent. Try it on your current project. Nothing you answer leaves your browser.

Make the Contact Test

Decision. Can you name the business decision this project makes better — in one sentence, without the word "platform"?

Owner. Is there a business owner whose own annual goals depend on the result — not an IT sponsor who inherited it?

Users. Have the people who must use it every day asked for it — or at least seen it before the rollout plan was written?

Trust. Would the project pass a data protection and governance review today — not one scheduled for just before launch?

Habits. Does the plan say out loud whose role, process, or daily routine changes — and who tells them?

A note on the numbers in the feedback: the 92 percent figure comes from the research behind my post The Confidence Gap. The developer study is a randomized controlled trial by METR, which I unpack in an upcoming post. The cancellation forecast is Gartner's. The test itself is deliberately blunt. So is reality in month six.

How I use it: before a budget decision, as a go or no-go conversation. In steering committees, when a project feels stuck and nobody can say why. And in talks and workshops, as the opening exercise — it is remarkable how quickly a room agrees on which force is the weak one.  If you want it applied to your organization, find me on LinkedIn or write me a mail.