Executive agent

A US executive search firm placing C-suite roles at Fortune 500s and senior roles at venture- and PE-backed companies.

Their competitive asset was a database of roughly 400,000 records, but exploiting it depended on a researcher knowing what to look for.

Every search started from scratch: read the client material, infer the role spec, decide on a strategy, then hand-assemble a pool from internal records and external sources that don't talk to each other.

The database covered a strong share of the addressable market for the roles they place, and most of that coverage was never reached in any given search.

A full-stack system, not a tool bolted onto the CRM. A data pipeline, an agent running in a custom harness, and a web interface with an interactive project model.

It takes client documents and call transcripts, with prior approval, and produces the role spec, the search strategy and a first candidate pool.

It searches the internal database and the open web in the same pass, including company websites, and reconciles the two.

It posts the pool to the ATS, absorbs the recruiter's feedback and runs again. Around the search itself it flags business-development openings, drafts outreach, runs background checks and surfaces back-channel references.

The agent sits at the centre of the firm's search model rather than beside it. In its first quarter in production, fifty consultants and researchers put it to work on live client engagements, running 11,300 jobs across the internal database and the open web — 99% of them to completion.

The pattern of use is the finding. Initial sourcing — the task it was built for — accounts for fewer than a quarter of those sessions. The rest is the work around a search: pipeline and market analysis, company research, candidate assessments, back-channel references, outreach, and new-business preparation. Given one instrument that reaches the whole record, the teams pointed it at the whole job.

A search now starts from a ranked pool assembled across the firm's full history and the open market in a single pass, with the reasoning attached, rather than from what a researcher happens to recall. Weekly use roughly tripled over the quarter, with no mandate behind it.