AI for recruitment agencies: 7 jobs to hand over, and what to keep
How recruitment and staffing agencies use AI in 2026: seven jobs worth handing over, what Bullhorn and JobAdder already do, the hiring rules, and a worked example.
Recruitment agencies get the most from AI on the work around a placement, not the placement itself: replying to every applicant within minutes, asking the basic screening questions, booking interviews across three diaries, reading CVs into the ATS, chasing documents, drafting job ads and shortlist summaries, and following up new client enquiries the same hour. Keep decisions about candidates with a consultant, because that is where the hiring rules in the US, the UK and the EU are strictest. Use the AI already in Bullhorn, JobAdder or your own CRM before buying more. Below are the seven jobs worth handing over, what your software already does, the rules in plain English, and a worked example for a ten-consultant agency. None of it is legal advice.
- Candidates notice silence more than anything else: in Greenhouse's October 2026 survey, 81% of US candidates had been turned down in the past year without speaking to anyone at the company.
- The fastest-growing agencies use AI most. Bullhorn found 78% of firms that grew revenue by more than 25% use AI built into their ATS.
- Hand AI the admin: replies, questions, booking, reminders, CV parsing, chasing and first drafts.
- Keep ranking and rejection with a person. That's where New York, Illinois, Colorado, the UK and the EU all set conditions.
- Whatever you add must write back to your CRM. A second database is worse than none.
How are recruitment agencies using AI in 2026?
Widely, but not deeply. Bullhorn's 2026 GRID report, a survey of nearly 2,300 recruitment professionals across North America, the UK and Ireland, Europe and Asia-Pacific, found that 46% of firms say AI has cut their screening time in half or better, and that among firms growing revenue by more than 25%, 78% use AI built into their applicant tracking system. Only 10% have AI agents working across the whole workflow, held back by worries about data, security and where to start.
Candidates, meanwhile, are still waiting by the phone. Greenhouse surveyed more than 2,800 job seekers in the US, UK, Ireland, Germany and Australia in 2026. Four in five US candidates (81%) had been rejected in the past year without ever speaking to anyone at the company, human or AI, and the most common complaint from those who did interview was being ghosted afterwards (45%).
FAST FOR THE AGENCY, SILENT FOR THE CANDIDATE
Agencies screen faster with AI. Candidates still hear nothing.
- Fastest-growing agencies (revenue up 25%+) using AI built into their ATS78%
- Agencies saying AI cut screening time in half or better46%
- Agencies with AI agents across the whole workflow10%
- US candidates rejected in the past year without speaking to anyone81%
The candidate’s side: silence, not speed, is what they remember.
- Interviewed candidates ghosted by the employer afterwards45%
Show the data as a table
| Measure | Share |
|---|---|
| Fastest-growing agencies (revenue up 25%+) using AI built into their ATS | 78% |
| Agencies saying AI cut screening time in half or better | 46% |
| Agencies with AI agents across the whole workflow | 10% |
| US candidates rejected in the past year without speaking to anyone | 81% |
| Interviewed candidates ghosted by the employer afterwards | 45% |
The gap between those two pictures is the opportunity for a small agency. Corporate hiring teams are buying AI that filters people out. An agency that uses AI to answer every candidate quickly and politely, and keeps a consultant on every decision, stands out to candidates and clients alike.
7 jobs worth handing to AI in a recruitment agency
| Job | What AI does | What the consultant still does | Where we'd build it |
|---|---|---|---|
| 1. First reply to every applicant | Replies within minutes on email, text or WhatsApp, answers questions from the job spec, asks availability, notice period, pay expectations and right to work | Reads the answers and decides who to call | AI agents |
| 2. Interview scheduling | Offers times from consultant and client calendars, books, sends the link or address, reminds, rebooks | Prepares the candidate and the client | AI automation |
| 3. CV and form parsing | Reads CVs and application forms into the right ATS fields, flags missing documents | Checks anything flagged | AI document processing |
| 4. Chasing | Chases references, right-to-work documents, timesheets and signed terms, politely and on schedule | Steps in when someone goes quiet twice | AI automation |
| 5. First drafts | Drafts job ads, shortlist summaries, candidate profiles and client updates from the consultant's notes | Edits and sends under their own name | AI integration |
| 6. New client enquiries | Replies to employers who enquire through the website the same hour, takes the role details and books a call | Wins the business | AI lead generation |
| 7. A searchable memory | Answers "who did we place in a similar role?" or "what did this client say about salary?" from your own notes, with the source | Decides what to do with it | Private knowledge base |
What's missing from the table is deliberate: ranking candidates, rejecting them, scoring interviews or reading their faces on video. Enterprise vendors sell all of those. For a small agency they carry the most legal weight and save the least time, because a good consultant can read a shortlist of 20 CVs faster than they can audit a tool that ranked them.
Start with the first reply
The first reply is where most agencies lose candidates, and it is the job that suits AI best. A good candidate applies to several roles in an evening. Whoever answers first, with something useful, usually gets the call back.
SUNDAY · ONE APPLICANT
AFTER HOURS
- 21:30WhatsApp: “Is the warehouse job in Dartford still open? I can start Monday.”
- 21:30“Hi, I’m the agency’s assistant, not a person. Yes, it’s open: days, £13.20 an hour, forklift preferred. A consultant can call you if you’d rather.”
- 21:32Asks about forklift licence, right to work, notice period and travel. Notes each answer, judges none.
- 21:34The candidate sends a CV photo. It is read into the CRM and attached to the job.
- 21:35Offers three interview times from the consultant’s and the client’s calendars. The candidate picks Wednesday 10:00.
- 21:36Sends the address, what to bring and a reminder for Tuesday evening.
IN THE CRM · MONDAY 08:00
CANDIDATE NOTE
- JOB
- Warehouse operative, Dartford
- LICENCE
- Counterbalance, expired 2025
- RIGHT TO WORK
- Says yes. Documents to check
- NOTICE
- None, available Monday
- INTERVIEW
- Wed 10:00 at the client
- NOT DONE
- Not ranked, scored or turned away. Consultant decides
Three rules make this work. The assistant says it is an assistant in its first message and offers a person. It answers only from the job spec and your own policies, so it never invents a salary, a shift pattern or a client's name you haven't approved. And it collects facts rather than judging them: a candidate without a forklift licence is noted as such, not turned away, because the consultant may know of a client who trains.
What Bullhorn, JobAdder and Recruit CRM already do
Check what you already pay for before buying anything. Bullhorn sells AI "digital workers" under its Amplify brand that enrich records, match, screen, send outreach, present candidates, transcribe and verify. JobAdder's Smart Match suggests candidates from your own database for each job. Recruit CRM, as of October 2026, includes AI CV parsing, candidate matching, job description and candidate summary generation, call transcripts and automated email sequences.
What these built-in tools usually don't do is the joined-up job: answer a candidate on WhatsApp at 21:30, check two calendars, book the interview, update the job in the CRM and tell the consultant in the morning. That's the gap a done-for-you setup fills, using your ATS's own API so everything lands where your team already works. If you run a less common CRM, or a spreadsheet, the same job can be built around it.
Worked example: a ten-consultant agency
These are assumptions, not measurements. Change them to match your own agency.
The agency: ten consultants placing warehouse, driving and office staff, 400 applications a week, 40 interviews a week, Bullhorn as the CRM.
Before, per week:
- Acknowledging and first-screening 400 applicants at 3 minutes each: 20 hours
- Arranging 40 interviews, with 15 minutes of back-and-forth each: 10 hours
- Keying 150 CVs into the CRM at 4 minutes each: 10 hours
- Chasing 60 people for documents and references at 5 minutes each: 5 hours
That's 45 hours a week, more than one full-time person, spread across ten consultants who'd rather be on the phone to clients.
After, consultants spend a quarter of that time reading what the assistant collected, checking flagged CVs and handling the exceptions: about 11 hours. The agency gets back roughly 34 hours a week, and every applicant hears back in minutes rather than days.
What it might be worth: if the time goes into business development and wins one extra permanent placement a quarter, at a 20% fee on a $45,000 salary that's $9,000 a quarter. Your fee and salaries will differ, and temporary placements are counted differently, so run your own numbers before believing anyone's, including ours.
The hiring rules, in plain English
Hiring is the most regulated place to use AI in a small business. The rules below are summarised as of November 2026. This is general information, not legal advice, and the place that matters is where the candidate is, not where your office is.
New York City. Local Law 144, enforced since 5 July 2023, says an employer or employment agency may only use an automated employment decision tool if it has had an independent bias audit within the past year, a summary of the results is published, and candidates are told at least 10 business days before it is used.
Illinois. Since 1 January 2026, the Illinois Human Rights Act (as amended by HB 3773) requires employers to tell people when AI is used in recruitment, hiring and other employment decisions, and forbids using it in a way that discriminates.
Colorado. Colorado rewrote its 2024 AI Act. SB 26-189, signed on 14 May 2026, applies from 1 January 2027 to technology that ranks, scores or recommends in a decision about an individual, including hiring. It requires notice before such a decision and, after an adverse one, a plain-language explanation and the chance of human review.
Across the US. In Mobley v Workday, a federal court in California ruled in July 2024 that claims could proceed against the software vendor on the basis that a tool which rejects or advances candidates is acting as the employer's agent. The case is still going. The lesson for agencies is that "the software decided" is not a defence for anyone in the chain.
United Kingdom. The Data (Use and Access) Act 2025 rewrote the UK GDPR's rules on automated decisions; the changes took effect on 5 February 2026. In March 2026 the ICO said employers can use solely automated decisions in hiring if they have a valid reason, but must tell candidates, explain how the tool works, let them contest a decision and ask a person to look again, and test for bias. Its 2024 audit of recruitment AI vendors found tools collecting far more personal data than needed and keeping it indefinitely.
European Union. The AI Act lists AI used to recruit and select people as high-risk. Emotion recognition at work has been banned since February 2025. After the Digital Omnibus agreed in 2026, the high-risk obligations for recruitment tools apply from 2 December 2027.
The common thread is easy to follow. If a tool only answers, books, collects and drafts, and a person decides, most of these rules either don't apply or are easy to meet. The moment a tool ranks or rejects people, you need audits, notices and a human review process. That's why we build the first kind and talk clients out of the second.
Will AI replace recruiters?
No, but it is changing what a recruiter's day is for. The parts clients pay for are judgement, relationships and closing: knowing which candidate will stay, persuading them to take the call, and getting both sides to yes. AI does none of that well. What it replaces is the hour after lunch spent sending "thanks for applying" messages and chasing a reference. Bullhorn's survey points the same way: the fastest-growing firms use AI most, and 56% of them place people in under ten days on average.
What we would do
For most agencies with between 1 and 50 consultants:
- List the week's admin, using the worked example above as a template, and find the two biggest numbers.
- Switch on what your ATS already has, such as CV parsing and email sequences, and train the team to use it.
- Add the first reply and interview booking, connected to your CRM and calendars, with a clear handover to a consultant. Read every conversation for the first two weeks.
- Write down where AI stays out: no automated ranking, rejection or video analysis, and what candidates are told. Our AI consultancy can write that policy with you.
- Then add the rest, one job at a time: chasing, drafts, client enquiries, a searchable memory.
That's how we build AI agents for agencies: on top of the software you already run, with a person on every decision about a candidate. If you'd like the same thinking for other trades, we've written guides for accountants, law firms and estate agents and realtors. For which general assistant to give your consultants, see our comparison of ChatGPT, Claude, Gemini, Grok and Muse.
Next Step
Talk to us about AI for your agency
A free 20-minute call: which jobs to hand over, which to keep, and a written quote if it is worth doing.
Frequently Asked Questions
How can AI be used in a recruitment agency?
Mostly on the admin around hiring: replying to every applicant within minutes, asking the basic screening questions, booking interviews across candidate, consultant and client calendars, sending reminders, reading CVs into the ATS, chasing documents, drafting job ads and shortlist summaries for a consultant to approve, and following up new client enquiries. Decisions about which candidate goes forward stay with a consultant.
Which AI tool is best for recruitment agencies?
There isn't one. Start with the AI already in your ATS or CRM (Bullhorn, JobAdder, Vincere and Recruit CRM all include some), add a general assistant on a business plan for drafting, and only then add tools for the gaps, usually candidate messaging and interview scheduling connected to your own calendar and CRM. A tool that doesn't write back to your CRM creates a second database nobody trusts.
Is it legal to use AI to screen CVs?
Often yes, with conditions that depend on where the candidate is. New York City requires a yearly independent bias audit, a published summary and 10 business days' notice before an automated tool is used. Illinois has required notice since 1 January 2026, and Colorado's notice rules start on 1 January 2027. In the UK, candidates must be told about solely automated decisions and can ask a person to look again. In the EU, recruitment AI is high-risk, with the full rules applying from 2 December 2027. This is general information, not legal advice.
Will AI replace recruiters?
Not the parts clients pay agencies for: knowing the market, judging people, persuading a candidate and closing a placement. It is replacing the acknowledgements, chasing, diary juggling and keying that fill a recruiter's day. Bullhorn's 2026 survey found the fastest-growing agencies were the heaviest AI users, with the same teams placing more people, not smaller teams.
Can AI schedule interviews for a small agency?
Yes, without enterprise software. An assistant can offer the candidate times from the consultant's and client's calendars, book the slot, send the video link or address, remind everyone the day before and log it against the job in the CRM. If the client wants to move it, the assistant rebooks and tells the candidate. The consultant only steps in when something unusual happens.
Should candidates be told they are talking to AI?
Yes. Say so in the first message, give a way to reach a person, and never let the assistant pretend to be a named consultant. Several hiring rules require notice when AI is involved anyway, and candidates are more forgiving of a clearly labelled assistant than of a fake human who gets something wrong.
Can AI reject candidates automatically?
We advise against it. Automated rejection is where the rules bite hardest: bias audits in New York, notices in Illinois and Colorado, the right to human review in the UK, and high-risk duties in the EU. In Mobley v Workday, a US court let claims proceed on the basis that a screening tool can act as the employer's agent. Let AI answer, collect, book and summarise, and let a consultant decide.
How much does AI cost for a recruitment agency?
The AI in your ATS may already be in your plan or a paid add-on, and general assistants cost roughly $20 to $30 per user a month on business plans. A done-for-you setup that connects candidate messaging, calendars and the CRM is a one-off build plus small running costs, quoted in writing after a call.
Next Action
Want help implementing this in your business?
Tamkoz builds AI phone receptionists, bespoke chatbots, and business automations. All work is quoted in writing before starting.