On this page
Recruiters already use AI to listen and write: call notes, summaries, briefs and first drafts. Few let it decide who to put forward, and they’re right not to, because that opinion is what clients pay for. The biggest gain left is on the client side: chasing a hiring manager who has gone quiet, and taking a stalled shortlist to another company that’s hiring.
What recruiters told us
We asked agency recruiters, on LinkedIn and on calls, what they run their desks on and what still costs them time. Most of them recruit in tech. Three things kept coming up.
- It listens
- The AI they named most was for notes: Granola, Fireflies and SourceWhale on calls, Claude and Copilot to turn notes into bullet points. One called automatic call notes a game changer, because without note-taking the conversations are better.
- They decide
- “Clients are paying us for our opinion.” The cover sheet, and the call on who goes forward, stay with the recruiter. So does the counter-offer conversation, handled person to person.
- Clients stall
- Clients’ time came up more often than finding people. A placement died at final stage when the client went quiet for two weeks, then came back with a lower offer.
What AI recruitment tools do today
Most AI recruitment tools fall into a handful of types. Nearly all of them work on the candidate side of the desk.
| Type | What it does | Used by recruiters we spoke to |
|---|---|---|
| Note-takers | Records and transcribes calls, then writes the summary | Granola, Fireflies, SourceWhale’s call notes |
| Assistants | Turns notes into bullet points, drafts emails and briefs | Claude, Microsoft Copilot |
| Sourcing | Searches for candidates and sends outreach sequences | Pin, SourceWhale |
| Screening and AI interviews | Scores applicants or runs a first interview | None named |
| AI in your CRM or ATS | Summaries, matching and automations inside the system you already useAgency CRMs are adding more of this | Ashby’s recorded, transcribed screening calls |
| Agency desks | Works each role from brief to placement, including the client side | Hamrr, which is us |
Why clients are the scarce side now
In the KPMG and REC Report on Jobs, candidate availability has risen for three and a half years, while permanent placements shrank every month for 45 months. They have only just started rising again. In the UK this year, for every 10 recruitment agencies that opened, 11 closed. See the numbers.
So the scarce side of an agency desk is now the client. Unless you recruit in a niche where the people really are scarce, the work that wins or loses a fee now happens after the shortlist. That’s where AI can do the most for an agency, and where it’s least used.
The loop: from your desk to Hunt and back
Every role runs through the same six steps, and the last one feeds back in: a stalled shortlist, taken to the right company, comes back to your desk as a submission to a new client. At each step AI can do the paperwork, and you keep the decision.
- BriefAI pulls the must-haves.
- ShortlistAI checks each CV.
- SubmissionAI drafts and sends the email.
- Follow-throughAI drafts and sends the chase.Client quiet a week
- HuntAI finds who’s hiring. You approve every send.
- New clientSame shortlist, resubmitted
Brief
AI is good at pulling structure out of a messy call: the must-haves, salary, location, interview steps and start date. It can also show what the brief leaves out. Which must-haves the client will bend on is still yours to find out on the call.
Here are my notes from a call with a client about a new role. List the must-haves, the nice-to-haves, the salary, location, interview steps and start date, using only what's in the notes. Then list what the notes don't cover that I should ask the client. Don't fill any gaps with guesses. [Paste your notes, with the client's name taken out]
Check a brief for gaps with the free job brief checker, or read how to take a job brief.
In Hamrr: the role is set up from your notes and the client’s own words, and its must-haves carry through every step after it.
Shortlist
AI can read each CV and your call notes against the must-haves and show where each one is backed up, and where it isn’t. It can’t tell you who will interview well, who will take a counter-offer, or who the client will like. That’s the opinion.
A CV is personal data. Use a business account or a tool whose terms say your data isn’t used to train the model, and share only what the task needs. The free CV anonymiser uses AI on those terms to take out names and contact details, so the CV is safe to send to a client.
In Hamrr: each CV and call note is read against the role, every requirement is marked backed up, doesn’t match or not discussed, and a client-safe CV and cover sheet are made in your format.
Submission
The email that goes with the CVs is where AI saves time and where it can do harm. Let it lay out the facts. The two lines that say why you rate this person should be yours, because that’s what the client reads.
Draft a short email to a hiring manager submitting the candidates below for [role]. For each, give one line on how they meet each must-have, using only my notes. Leave a gap marked [WHY I RATE THEM] for me to write. Keep it under 150 words and don't add praise. [Paste your notes on each candidate, without names or contact details]
For the rest of the email, read what to send with a CV.
In Hamrr: the submission is drafted from the shortlist, with the client-safe CVs attached, and sent from your own mailbox when you say so.
Follow-through
This is where fees die quietly. The hiring manager goes on holiday, a counter-offer lands, the candidate takes another interview. AI can read the thread and notice the silence before you do, and draft a chase that asks a real question instead of “any update?”.
Here is my email thread with a hiring manager who hasn't replied for [number] days about a shortlist. Draft a short chase that names one candidate and the requirement they're most likely weighing, offers to find out more, and gives a true reason for the date I need an answer by. No apologies for chasing. [Paste the thread]
More on this step: chasing interview feedback, a client gone quiet on your shortlist and handling a counter-offer.
In Hamrr: your inbox is read for you. A week of silence from a hiring manager gets a drafted chase, and offers and counter-offers are flagged on the candidate.
Hunt: when the client stays quiet
A spec CV, sent to a company that didn’t ask for one, is how recruiters have always won clients. One recruiter told us spec CVs had brought them three new clients in a year and a half. It’s also the first thing to go when the desk is busy, because the research takes longer than the email.
That research is what AI is good at: past clients, companies with live ads for the same role, fresh funding, a new head of the team, and the candidates’ former employers. You still choose the candidate and the company, and you ask the candidate first.
Open your last five shortlists. How many are still waiting on a client? Each one is a spec email you haven’t sent.
Below is a real Hunt run, recorded and sped up. Pick a role whose client has gone quiet.
Loading the Hunt demo…
Next, read how to send a spec CV and business development without a BD team.
New client, same shortlist
A reply to a spec opens a new account. Agree terms, confirm the brief, and the same shortlist goes back to submission, drafted for you. The candidates you spent weeks on get a second shot, and you have a client you didn’t have before.
How recruitment agencies should choose an AI tool
Four questions to ask before you buy, whichever tool it is:
Does it work inside your inbox and ATS?
A tool with its own tab and login gets dropped within a month. What it learns should land in your records too: recruiters told us about typing the same candidate into their CRM and then into each client’s system.
Does it ask before it sends?
Anything that emails a candidate or a client goes out in your name. You should be able to see it first, and choose what can go on its own.
Where does candidate data go?
Ask where CVs are stored, who can see them, and whether they’re used to train a model.
Does it replace a tool, or add one?
Desks are already full. One recruiter runs six tools alongside the CRM, and several said the problem we asked about was handled by a tool they already had.
The whole loop, with your yes on every step.
Hamrr works each role from brief to placement on top of the inbox and ATS you already use, and Hunt takes a stalled shortlist to companies that are hiring.
- Works with Gmail, Outlook and 11 applicant tracking systems
- Asks before anything is sent, or switch a category to Auto
- Writes outcomes back to your ATS
Questions
Will AI replace recruiters?
Not the part clients pay for. The recruiters we spoke to use AI to take notes and draft, and keep the judgement, who to put forward and why, for themselves. AI is better used to stop that judgement going to waste.
What AI tools do recruiters use?
The ones the recruiters we spoke to named most were note-takers and assistants: Granola, Fireflies, SourceWhale, Claude and Microsoft Copilot, used alongside a CRM or ATS such as Bullhorn or Loxo.
What is the best AI tool for a small recruitment agency?
The one aimed at the work that loses you fees, not just minutes. A note-taker saves time on every call, but fees are lost after the shortlist, while a client is quiet. Pick a tool that works inside your inbox and ATS, asks before it sends, and replaces something you already pay for.
Is it safe to put a CV into ChatGPT or Claude?
A CV is personal data. Take the name and contact details out first, use a business account whose terms say your data isn't used to train the model, and share only what the task needs.
How can AI help with recruitment business development?
It can do the research a spec CV needs: which companies are hiring for the same profile, why now, and who to write to. Then it drafts the email. Which candidate, which company and when to send stay with you.