Case study · Industrial components & equipment · Switzerland
A real engagement, client anonymised. Figures measured with the client and rounded; not audited by a third party.
A 240-person industrial group goes from 43 AI ideas to 3 funded projects, 2 of them now live
In three months, 43 ideas and three isolated trials become three funded projects, each with an owner and a starting measurement; two are live nine months later.
Scope and limitations
Rounded figures, recorded at the end of the engagement and at 9 months for go-live. A real engagement: the client’s name and any detail that would identify it have been removed.
- Company
- Swiss industrial group (components and equipment), around 240 people, 2 sites in French- and German-speaking Switzerland
- Markets
- Export customers; part of the customer data is a trade secret
- Software in place
- Group ERP and Microsoft 365 on both sites; an IT team of 4
- Starting point
- 43 ideas from internal workshops, 3 isolated trials, none live
Starting situation
At every meeting, the board asked the same question: what are we doing with AI? The company had answered with numbers: internal workshops had produced 43 ideas, three trials had been started in different departments, and several vendors were each pushing their own demo. None of the trials had gone live.
The problem was not a lack of ideas but the lack of criteria to choose. Taken on its own, each idea looked reasonable; none had an owner, a starting measurement or a condition for stopping. The annual budget was approaching: without a reasoned decision, the subject would slip a year or be funded blind. Management had been debating it for eight months.
The question asked
Where can AI deliver a real gain first, and what has to be settled before investing?
Constraints
Trade secrets
Technical and commercial data of export customers is covered by confidentiality obligations. Every idea had to state where that data goes before it was even assessed.
Production at full load
Production teams had no time to spare. Any idea that needed their time had to justify it.
An IT team of 4
IT already ran the ERP and the infrastructure of both sites. Each idea was judged on what it costs to run, not only to build.
The budget calendar
The decision had to be ready for the annual budget. An analysis delivered too late would have been worthless.
What we put in place
The deliverable was not a list of ideas but a way of deciding: link each idea to a measurable gain, compare them on written criteria, check the state of the data, stage the investment in three steps, and repeat that choice every month, without us.
The building blocks
Stages of the engagement
Field work and task mapping 3 weeks
Workshops with the teams on both sites and direct observation in production. The 43 ideas are tied to real tasks; those with no identifiable gain are dropped straight away.
Scoring and feasibility 3 weeks
Gain × feasibility scoring of the remaining ideas, interviews with the future owners, checks on vendor dependencies and production constraints.
Data and architecture 4 weeks
State of the data needed by the leading projects (quality, access, sensitivity) and an architecture that fits the ERP, Microsoft 365 and an IT team of 4.
Decision and timetable 2 weeks
A single investment proposal: 3 projects to fund, three steps, decision points and follow-up. Presented to management, then to the board.
- Every project has a named business owner before it is selected
- A starting measurement before any funding: no estimated gain without a point of comparison
- Written decision points: the right to try is not the right to extend
- Written criteria for choosing between building, buying or adapting
- Monthly check-in: continue, redirect or stop
- No trial without a stopping criterion set before it starts
Measured results
An engagement like this is judged twice: on the decision it makes possible, then on what actually exists a few months later. The figures compare the situation before and after the engagement; go-live is measured at 9 months.
Rounded figures, recorded at the end of the engagement and at 9 months for go-live. A real engagement: the client’s name and any detail that would identify it have been removed.
What we would do differently
Every study also says what could have been done better.
- 01
Go to the shop floor in the first week. The most valuable project came from a conversation with a team leader at his workstation, not from the ideas workshops: field observation should have come first.
- 02
Simplify the scoring grid. Management decided on 3 criteria while the grid had 9: the other six documented the analysis but slowed every discussion.
- 03
Involve the finance director from the start. The cost calculation had to be redone to fit the budget format, a detail that belongs at the start of an engagement, not the end.