Deciding whether to build
The year generative AI moves from technology watch into meetings, where the question is often badly framed. And the year my role in those meetings changed.
Two changes in the same year. Generative AI reaches committees, where I am asked less to make it work than to say whether it should be used. And the setting I work in changes. Both point to the same reading: books that help prepare a decision.

Its brevity is its main merit: it is an overview you read in full, which beats a treatise abandoned halfway. It is enough to know what you will not manage without help.

A lesson in method more than in economics: what can be claimed about the effect of an intervention, and at the cost of what experiment. The transfer to business is direct, and almost never made.

The book furthest from the trade, and the most useful of the year. Its thesis, that ancient philosophy was a daily exercise before it was a body of texts, applies well to a trade learned through practice more than through theory.