How Generative Talent Co found the AI founders before the role existed
A confidential search firm with no list to buy, no logos to show and no sales team. We built the shape of their buyer from evidence, watched the AI market for the moment it changed, and approached on two channels as one conversation.
A search firm selling something nobody has a budget line for yet.
Generative Talent Co recruit for AI companies. Not the roles that have existed for twenty years, but the ones that turn up for the first time when a company decides to stand up a new discipline. An agentic workflow architect. A governance lead. A first commercial hire into a research team. Roles with no title history, no salary benchmark and nobody internally who has hired one before.
That is a good business to be in and a hard one to sell into. The firms who need them have usually not written the job description yet, and often have not admitted to themselves that they need to. By the time a role is posted publicly, three other firms are already on it.
There was a second problem on top. The work is confidential. Client names are withheld, outreach goes out under an anonymised description, and shortlists are handed over behind a link and a six digit code. So there are no logos to put on a deck and no placements to name. The usual proof a search firm leads with was not available.
Four reasons the usual approach was never going to work.
Build the shape of the buyer, then wait for the market to move.
We started with what Generative Talent Co actually sell and who they have already delivered for, rather than a whiteboard session about the ideal customer. From that we narrowed every company that calls itself an AI business down to the shape that converts for them, and built the account list from that shape rather than from anything bought off a shelf.
A static list goes stale in a month. Sonar keeps the account list under continuous watch and only puts a company in front of a person when something inside it moves. The approach is timed to the change, not to a sending schedule.
Most firms run these as two separate campaigns and the prospect gets two unrelated messages from two strangers. Here they are a single sequence against the same account, written from what the signal says about that company, and referring to the change rather than opening with the service.
Replies are graded rather than counted. Interest, referral, not now, no. That grading feeds straight back into which companies are worth approaching next, so the list gets better the longer it runs instead of being rebuilt from scratch each quarter.
