A recruitment agency runs on email, and most of it is logistics
An agency recruiter has two relationships to maintain and one inbox to do it in. Candidates want to know where they stand, clients want to know who is coming, and both go quiet at inconvenient moments. The work that fills the day is chasing, confirming, and repeating information that already exists somewhere.
4 min read
Two agents, not one
Start here, because it is the decision that prevents the worst failure. The candidate side and the client side are different conversations with different confidentiality, and one agent holding both can quote one to the other.
Two identities, two keys, no cross grants, per running several email agents in one workspace. The isolation is not paranoia: a candidate's salary expectation reaching a client, or a client's feedback about a candidate reaching that candidate, is the kind of mistake that ends a placement and sometimes a relationship.
The candidate side
Acknowledge every application, immediately. The single most common complaint about recruiters is silence, and it is entirely solvable. An acknowledgement that says what happens next and when is worth more than a personalised rejection three weeks later.
Keep people informed at each stage. Submitted to the client, interview scheduled, feedback pending. Short, specific, and honest about timelines that have slipped.
Collect what the process needs. CV updates, right to work documents, references, availability. This is where placements stall, and an agent that asks precisely and tracks what is outstanding removes real drag, per attachments.
Coordinate interviews. The scheduling loop, which is the same one as everywhere else, per automating interview scheduling email.
What it must never do: reject anyone, discuss salary beyond restating a published range, characterise a client or another candidate, or answer why someone was not progressed. Those are conversations that need a person and occasionally a careful one.
The client side
| Task | Agent | Person |
|---|---|---|
| Confirming a shortlist has been sent | Yes | |
| Chasing interview feedback | Yes, twice, then escalate | |
| Scheduling with the hiring manager | Yes | |
| Sending candidate profiles | Yes, from the record | |
| Negotiating fees or terms | Always | |
| Pushing back on a brief | Always | |
| Anything about a placement falling through | Always |
Chasing feedback is the highest-value item on that list. Clients go quiet after interviews, candidates read the silence as rejection and accept elsewhere, and a polite chase on day two rather than day nine is the difference between a placement and a wasted month.
Memory is the compounding advantage
Recruiters carry an enormous amount in their heads: who is looking, who is not right now but will be in spring, which client hates being called, which candidate is worth a call the moment a particular role appears.
Some of that belongs on the contact record, and the agency that records it systematically has an asset when someone leaves. Preferences, constraints, what a candidate is looking for, what a client's process actually looks like in practice, per giving an email agent memory.
What does not belong there is judgement about people. "Prefers remote, available from September" is a fact. Anything evaluative about a candidate is a record you may one day have to produce, and a data subject access request is not the moment to discover what your agent wrote, per the compliance floor.
Sourcing is where the rules bite
Outbound to candidates who have not applied is cold contact, with everything that implies: it needs a lawful basis you can state, an opt-out that works, and volume low enough to be defensible. An agent makes it easy to do this badly at scale, which is precisely why the limits sit under the tool rather than in the prompt, per keeping an autonomous agent from becoming a spam problem.
The version that works is small and specific: a handful of genuinely relevant people, a message that could only have been written for them, and no chasing after the second attempt.
Where to start
The candidate acknowledgement and status updates. It is the highest-volume, lowest-risk category, it fixes the complaint your candidates actually have, and the worst realistic failure is a slightly stiff message rather than a lost client. Widen to the client side once you have read a fortnight of threads, per the first week with an email agent.
Questions
- Should one agent handle both candidates and clients?
- No. Two identities with no cross grants, because one agent holding both sides can quote a candidate's expectations to a client or a client's feedback to a candidate.
- What is the highest-value task?
- Chasing interview feedback on day two rather than day nine, and acknowledging every application immediately. Both are silence problems, and silence is what candidates complain about.
- What must the agent never do?
- Reject anyone, negotiate fees or salary, characterise a candidate or client, or explain why someone was not progressed.
- Can it write candidate notes?
- Facts, yes: availability, preferences, constraints. Not evaluations. Those become records you may have to produce in a data subject access request.
- Is candidate sourcing safe to automate?
- Only carefully. It is cold contact, so it needs a lawful basis, a working opt-out, and small, specific volume. Scale is what turns it into a problem.
- What should it collect?
- Documents and details the process needs, tracked so it knows what is still outstanding. That chasing is where placements stall.
Give your agent an address it can answer from.
Create an inbox