The Contract Problem

A founder rang me a couple of months ago in a bit of a state. One of her biggest clients had sent over a new AI addendum to their master services agreement and asked her to sign it within two weeks.

She hadn't read it yet, but she'd glanced at it and the gist was that the client wanted to know exactly which AI tools were being used on their account, what data was being processed, and what the agency's internal policies were on all of it.

She didn't have good answers to any of those questions. Not because they were doing anything wrong, but because nobody had ever asked. The tools had been adopted gradually, team by team, and the governance hadn't kept pace. When she looked at her own standard contract, it still referenced data processing in terms that predated the existence of ChatGPT.

What surprised her, and this is something I'm hearing more and more, was that the client wasn't hostile about it. They were going through their own AI governance process and simply needed their suppliers to be at the same level. It was a procurement exercise, not a confrontation.

She ended up treating it as the forcing function she'd needed. Wrote a proper AI use policy, mapped which tools were used where, and updated her own contracts so that every new engagement had an AI processing clause built in from the start. It took about three weeks.

The agencies that will find this easiest are the ones who do it before they're asked. The frameworks are at scaleatspeed.com/ai if you want a starting point.