The context
A major industrial group wanted to know where AI could create value in its operations. Among the pain points raised by the teams, one came up again and again: producing operational reports, consolidated by hand from several tools.
The approach
- Understand. Activity analysis and interviews with the teams involved.
- Prioritize. A workshop with stakeholders to rank use cases by value and feasibility.
- Plan. A phased AI roadmap, presented to leadership.
- Build. For the first use case: functional and technical scoping, data source analysis, prototype, then go-live.
- Transfer. Technical documentation, user guide and team training.
First system: operational reporting
- Over 70% of time saved on producing operational reports.
- Fewer copy errors.
- A documented system, owned by the business team, then extended to other reports.
Second system: technical reports
Under the same roadmap, a second use case covered technical reports produced from simulation data. A critical, regulated activity: the process took a week and involved four people at different stages.
- A workflow redesigned in eight steps. AI drafts from the simulations, the engineer checks and signs.
- A system on the client’s servers. The model was chosen according to data classification, and a no-AI mode lets work continue by hand.
- Half a day for one engineer instead of a week for four people, based on user feedback. The users were trained, and first-level maintenance was handed to an in-house engineer.
The right first project is not the most spectacular one. It is the one where the problem, the data and the owner are already clear.
On video
We discuss this project in an interview with Malt.
What we take away
Reporting was a good starting point: a measurable gain, data already being produced and a team ready to use the result. This first outcome gave leadership concrete evidence for the rest of the roadmap.
Is a manual process costing you time? Let’s talk for 30 minutes.
Book a callMore notes