Insights
Working notes from the people doing the work.
We don’t write hot-take posts. The notes below come out of real engagements - the patterns we keep seeing, the mistakes we made and don’t intend to repeat, the things we wish we’d known six months earlier.
Latest
Why most agentic AI pilots stall at “interesting”
The pattern is well-rehearsed by now. Three months in, the prototype impresses in demos and misses on every metric anyone cares about. Why it happens, and what distinguishes the pilots that don’t stall.
Read the post →BPM isn’t dead, it just got useful again
Process management was unfashionable for most of the last decade. Agentic AI has quietly made it interesting again - because agents need workflows the way payments need ledgers.
Read the post →Evaluations are not optional
If you can’t grade your agent on a representative test set, you don’t have an AI system; you have a prompt that worked once. A practical guide to building an eval harness that pays for itself in the first month.
Read the post →Topics
Got something you’d like us to write about?
If there’s an agentic, AI or BPM question your team is wrestling with and you can’t find a sensible piece of writing on it, tell us. We’d rather write about real questions than invented ones.