Process and knowledge management
Learning from data
Your systems already know a lot about your workflows. How to read it, what to learn from it and where AI helps.
Traces in your systems
Every system you work with keeps a record: when an invoice arrives, when it is checked and when it is paid. Three details are enough to reconstruct the actual workflow from that: the case, the step and the time. This approach is called process mining.1
| Case | Step | Time |
|---|---|---|
| Invoice 2026-118 | received | 2 Mar, 08:14 |
| Invoice 2026-118 | checked on content | 9 Mar, 15:40 |
| Invoice 2026-118 | back to site management | 10 Mar, 11:02 |
| Invoice 2026-118 | checked on content | 16 Mar, 09:25 |
| Invoice 2026-118 | approved | 17 Mar, 13:10 |
| Invoice 2026-118 | paid | 24 Mar, 10:00 |
This one invoice already shows a loop and a week of waiting before the first check. Evaluated over hundreds of invoices, you see which path is the usual one and where the time is lost.
What you learn from it
| Question | What the data shows |
|---|---|
| How does it really run? | The actual workflow with all its detours, not the one from the manual |
| Does it run as agreed? | Where the lived workflow deviates from the agreed one, such as missing approvals |
| Where does it stick? | Waiting times, loops and bottlenecks, with figures instead of guesses |
Where AI helps
Artificial intelligence is useful where there is a lot of text and little structure: it finds the place in minutes, e-mails and on the knowledge platform where something has been solved before, reads details from delivery notes and invoices and summarises long threads. It does not replace the judgement of experts, it shortens the way there.