A practical playbook for using AI on real EU proposals — without handing your draft to a cloud service, and without pretending a machine can tell you whether you will be funded.

This second book picks up where the Grant Writing Book leaves off. That guide is about designing and structuring a research project. This one is about the work that follows: drafting under deadline, deciding where AI genuinely helps, and getting an honest second reading of your text before a panel gives you theirs.

It is built on something unusual: 71 real proposals, each paired with the official Evaluation Summary Report its evaluators wrote back. Most were accumulated over fifteen years of SNM consulting; others are public documents — evaluation reports released under EU access-to-documents rules or obtained through freedom-of-information requests. From those reports come 461 distinct reviewer criticisms and 462 recorded strengths: the actual sentences panels used, criterion by criterion. The book is organised as 18 recipes across six phases, from qualifying a call to submission, each grounded in what evaluators actually wrote rather than in what guidance documents say they want.

It is deliberately honest about limits. No tool predicts your score. What the evidence supports is narrower and more useful: knowing the objections real evaluators raise on passages like yours, while you can still do something about them.