Skip to content
FQ AI Summit 2026

Make Continuous Learning Compatible With Stable Work

My notes from attending the FQ AI Summit in New York, November 10, 2026.

AI Experiments and Institutional Knowledge

How do we keep improving without changing everyone's process every week?

The continuous innovation panel talked about learning as an everyday habit. The people using a service still need to know which process is current and dependable. Try new approaches somewhere separate and approved, and keep one clearly identified current version for everyone else.

Testing happens in its own space

Test new approaches in an approved space with clear boundaries. Don't replace a working instruction without telling anyone just because a new demonstration looks promising.

Pick review points for moving a change into the operating process. Say what evidence is required and who decides, and count the cost of retraining and updating connected material.

People need one clear current version

Tell affected colleagues what changes and what stays the same, so nobody has to choose between several plausible instructions. Urgent corrections can use a separate path when necessary, since stability shouldn't delay a fix that protects people from a known problem.