The cost of being the innovators

Last week I wrote about creating your own playbook for AI engineering. Teams should look at what others are doing, experiment internally, and build their own way of working.

Then a friend watched a video about an AI harness playbook and spent the whole week redesigning and rebuilding their harness and workflow, trying prompt structures and formalizing an internal process. They are a four-person company.

This has a cost: one hidden by the AI hype, and one you may pay again when the next thing comes out or someone publishes a new way of working on Hacker News.

Do not be an innovator if AI workflow design is not your core business. Experimenting is good. Turning your engineering team into a permanent AI strategy committee is not.

It is easier to be an early adopter, or part of the early majority, adopting concepts once they have been proven in similar environments. Yes, experiment, but do not confuse experimenting with endlessly searching for the perfect workflow.