Industrial manufacturing — 2026
Ninety AI ideas, twenty worth funding
An enterprise-wide AI discovery for a large manufacturer: pain points mapped across five value streams, use cases scored on impact and feasibility, and a roadmap tied to EBITDA.
Stack
The problem
A manufacturer with a long list of AI ideas and no way to tell them apart. The list had been assembled from vendor pitches, internal enthusiasm and a consultancy deck. Nobody could say which ones the company’s own data could actually support.
The method
Five value streams — procure-to-pay, order-to-cash, and three production and dispatch flows — worked through the same way:
- Pain-point inventory. What hurts today, who feels it, what it costs. In the operators’ words, not the transformation office’s.
- Data-readiness per pain point. Does the data exist, is it accessible, is it trustworthy, is it timely. Most ideas die here, and that is the point.
- Solution shaping. For surviving pain points, what class of solution fits — and often the answer is a rule, a dashboard, or a fixed process rather than a model.
- Impact × feasibility scoring, mapped onto the EBITDA value areas the finance function already recognises.
Alongside it, an AI maturity assessment covering governance, data foundations, skills and operating model — because a portfolio is only executable if the organisation behind it can run it.
Outcome
Roughly ninety candidate use cases reduced to twenty with a defensible business case, sequenced into waves, with the data work each wave depends on made explicit. The uncomfortable half of the output was as valuable as the funded half: a documented reason each rejected idea was rejected, which stops it coming back next quarter.