AI agents inside the search firm: honest notes from six months of production use.
We put agentic tools into live mandates in early 2026. What worked, what did not, and why the consultant conversation still decides every senior hire.
In January we made a deliberate decision to put agentic AI tools into our live mandate workflow — not as a marketing exercise, but as a real production experiment. Six months in, here is what we have learned, honestly.
The wins are real and specific. Longlist construction is 60% faster. Structured research on a target company — org design, recent leadership changes, likely internal candidates — is a task an agent now does in an hour that used to take a researcher a full day. Reference-check preparation is dramatically better; the agent surfaces context our consultants would have taken three calls to find.
The failures are also real. The agent cannot read a candidate. It cannot tell whether the operator across the table is running toward your mandate or away from their last one. It cannot hold a boardroom conversation about compensation optics. Every mandate we have tried to over-automate has stalled at the interview stage, because the humans on both sides need the human on our side.
The most useful pattern that has emerged is what we now call the two-thirds rule. Agents own the first two-thirds of the pipeline — market mapping, longlist, initial research, structured outreach templates. Consultants own the last third — calibration calls, candidate reads, committee management, offer choreography. When we cross that line in either direction, the mandate suffers.
One more note that surprised us. Candidates can tell within thirty seconds whether an outreach was written by a person or a template. The response rate on agent-drafted outreach is 40% lower than consultant-drafted outreach, even when the copy looks identical. Senior operators read subtext, and there is a subtext to a message that was not written for them.
Our view for the rest of 2026: the search firms that will win are the ones who use AI to compress the mechanical work and reinvest the saved hours into more candidate time, not fewer consultants. The firms that use AI to cut headcount will produce worse shortlists, and the market will notice within a year.
