Make AI fit real work, not the other way around.
I am an activity ergonomist. I help make work situations, constraints, responsibilities, and AI effects visible enough to design better delegation loops, tools, and organizational decisions.
About Julien Talbot
A technology never enters an org chart. It enters gestures, trade-offs, cooperation, and responsibility. That is where design has to start.
Three paths are enough.
The English surface is intentionally narrower: collaborate on AI/product reliability, invite a talk, or read the underlying thesis.
Build systems that fit real activity.
Evaluation, post-training, software products, and AI agents when delegation, evidence, and recovery become critical.
Create a shared language.
Talks and workshops on what AI changes in work, responsibility, supervision, and decision-making.
Read the core thesis.
Writing on real work, AI agents, delegation, evidence, and the conditions for trustworthy transformation.
Public stages, real work situations, production-adjacent agents.
The subject shifts with the audience: work organization, prevention, AI, software products, post-training. The criterion stays the same: make the real activity visible before deciding what to automate, transform, or delegate.
- 7
- talks and workshops in 2025/2026: UNEP, CCI, Sorbonne, Préventica, Cinov Ingénierie, Foire Agricole de Bras-Panon, French Tech Réunion.
- 24
- replay-ready, oracle-locked agent cases from recent work on tool-using agent reliability.
- 8
- daily-use axes: delegation, external state, memory, runtime, recovery, environment, files/code, blockers.
Start with the work situation
For talks, strategic discussions, or collaboration on AI agents and real work systems, email me directly. If the topic is still fuzzy, the guided request helps shape the useful context.