The short version: Recruitment agencies now use AI to parse CVs, rank candidates against a job spec, run first-stage chat or video screening, and auto-schedule the ones who pass, cutting shortlist time from days to hours. It works brilliantly for volume roles and badly for anything nuanced, and most agencies never audit who the AI quietly rejects. If you're hiring through an agency or running one, that gap is the thing worth understanding, not the speed stat on the sales deck.
What agencies mean when they say "AI screening"
When a recruitment agency tells a client they use AI to screen candidates, they usually mean one of four things happening in sequence: parsing (pulling structured data out of a messy CV), matching (scoring that data against a job spec), conversational screening (a chatbot asking pre-set questions before a human ever sees the application), and ranking (sorting hundreds of applicants into a shortlist a recruiter can work through in a morning instead of a week).
None of this is new in concept. Applicant tracking systems have used keyword matching since the 2000s. What changed around 2023 to 2025 is that the matching got smarter (large language models reading a CV in context rather than just counting keywords), and the conversational layer got good enough that candidates often can't tell they're talking to software for the first two or three exchanges.
The tools doing the actual work
Most UK and US agencies are running one of a handful of platforms, or a combination: Bullhorn and Manatal for core ATS and parsing, Textkernel for CV parsing specifically (it's the engine behind a lot of white-label ATS features you'd never know were Textkernel), Paradox's Olivia for conversational screening, HireVue for video interview analysis, and Eightfold or Beamery for the bigger enterprise-side talent matching. Smaller agencies increasingly bolt a custom GPT-based screening layer onto whatever ATS they already have, because building a simple screening chatbot is now a two-week job rather than a six-month one.
A real example: what this looked like at a 40-person Manchester agency
I worked with a recruitment agency last year that specialised in warehouse and logistics placements, roughly 40 staff, placing around 1,200 candidates a year across the North West. Before automation, each consultant was manually reading every CV that came in against six or seven open roles at a time. Their own numbers: an experienced consultant could get through about 15 to 20 CVs an hour if they were being careful, which meant a busy Monday with 300 new applications took the best part of two days just to triage.
We put a parsing and ranking layer in front of their ATS. The AI read every incoming CV, extracted relevant experience (forklift licence, shift patterns worked, distance from site), and scored it against the live job specs. Anything scoring above 70 percent went straight to a consultant's queue. Anything between 40 and 70 got a short automated screening chat, three questions, right to work status, licence expiry, notice period, before being routed. Anything under 40 was auto-archived with a polite rejection email.
Result after eight weeks: time to shortlist dropped from an average of 11 days to just under 2. Consultants were spending their time on calls and site visits instead of reading CVs. That part is good, and it's the part every case study you'll read stops at.
What the case studies don't tell you
Here's what we found when we manually re-read a sample of 200 auto-archived CVs from that first eight weeks, purely as an audit exercise: 14 of them should have gone through. A candidate who'd worked as a "materials handler" instead of "warehouse operative" got scored low because the parser didn't map the job title, despite having exactly the forklift certification the role needed. A candidate with a two-year gap for parental leave scored lower not because the AI penalised the gap directly, but because the gap meant less recent "relevant experience" in the scoring window it used.
Nobody at the agency had checked this before we did. Not because they didn't care, but because the whole point of the system was that they no longer had to look. That's the uncomfortable bit nobody selling AI screening tools puts on the homepage: the efficiency gain and the quality risk come from the exact same mechanism. The system is fast precisely because a human isn't checking its judgement, and a human not checking its judgement is precisely how good candidates disappear without anyone noticing. Agencies that adopt this without building in a manual audit step aren't being efficient, they're being unlucky in a way they haven't discovered yet.
The step-by-step: how a screening pipeline runs
- Step 1, ingestion: CVs come in via job boards, agency website forms, or referrals, and get pulled automatically into the ATS regardless of format (PDF, Word, LinkedIn export).
- Step 2, parsing: The AI extracts structured fields, job titles, dates, skills, qualifications, location, and normalises them against the agency's own taxonomy.
- Step 3, scoring: Each candidate gets a match score against the specific job spec, weighted for must-haves (a driving licence, a security clearance) versus nice-to-haves.
- Step 4, tiering: Candidates split into bands, typically top tier goes straight to a recruiter, mid tier goes to automated screening, bottom tier gets auto-rejected or held in a talent pool.
- Step 5, conversational screening: Mid-tier candidates get a chatbot or SMS-based screen asking three to six qualifying questions, usually taking under five minutes for the candidate.
- Step 6, human review: A consultant reviews the shortlist, listens to any recorded video answers, and picks who to put forward to the client.
- Step 7, feedback loop: The best agencies feed placement outcomes (who got hired, who performed well after six months) back into the scoring model, though in practice most agencies skip this step entirely because it's slower and less exciting than switching on the next feature.
Where this earns its keep
For high-volume, low-differentiation roles, warehouse, call centre, hospitality, driving, this works well. The job specs are clearcut, the qualifying criteria are objective (licence, clearance, shift availability), and there are hundreds of applicants for each role, so speed matters more than nuance. This is where I'd tell a client to automate first if I were advising them, and it's covered in more detail in this breakdown of what to automate first across 19 sectors, because the answer for a logistics agency is completely different to the answer for an executive search firm.
For senior, niche, or client-facing roles, it earns its keep far less. A CFO search or a specialist engineering placement usually has fewer than 30 applicants, and the differentiators are things a parser can't read, how someone handled a specific board conflict, whether their leadership style suits a founder-led business. Agencies that apply the same automated screening logic across every role level are the ones I see losing placements to boutique competitors who still read every application by hand for their top-tier roles.
The bias and legal question agencies keep quiet about
In the UK, the Equality Act 2010 still applies to automated decisions exactly as it applies to human ones, an AI system rejecting candidates on a pattern correlated with age, disability, or maternity leave is still discrimination even if no human made the individual call. The Information Commissioner's Office has published guidance specifically on AI and automated decision-making in recruitment, and it puts the burden squarely on the employer or agency, not the software vendor, to prove the system isn't discriminating. Most small and mid-size agencies I've spoken to have never run a bias audit on their screening tool. They've bought a product, switched it on, and trusted the vendor's marketing that it's "fair by design." That's a genuine liability sitting quietly in a lot of agency operations right now, and it's one of the specific problems worth fixing rather than assuming away, which is exactly the kind of gap covered in this piece on fixing screening and follow-up without losing good candidates.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
What candidates should do differently
If you're job hunting through an agency, assume your CV is read by software first. That means mirroring the exact job title language used in the advert (not a close synonym), spelling out qualifications in full rather than abbreviations the parser might not recognise, and keeping your most recent and most relevant experience near the top rather than buried under a chronological format that puts your oldest job last. It's not gaming the system, it's writing for the actual reader, which happens to be software before it's a person.
What good implementation looks like
The agencies getting this right share three habits. They keep a human reviewing a random sample of auto-rejected candidates every month, not just the ones who complain. They set the automation thresholds differently by role seniority rather than using one blanket cutoff score for every vacancy. And they tell candidates plainly that part of the process is automated, which, counterintuitively, tends to increase completion rates on screening chats rather than putting people off. Candidates would rather know than guess.
If you're weighing up whether to build this in-house or bring someone in to set it up, it's worth looking at what's landing across other functions right now, not just recruitment, in this roundup of real-world AI agents working in businesses in 2026, because the same mistakes (no audit, no feedback loop, one-size-fits-all thresholds) show up everywhere, not just in hiring.
And if the honest answer is that nobody internally has the time or the technical grounding to set the bias checks and scoring logic up correctly, that's a fair reason to bring in outside help rather than a reason to skip the checks, which is where a proper look at what AI consulting costs is worth ten minutes of your time before you sign anything.
Related: ai consultant vs ai agency small business.
Related: automation: guidelines and how to pitch.
Frequently asked questions
Do AI screening tools discriminate against candidates?
They can, and often do without anyone realising, because the discrimination shows up as a pattern (older candidates, career-break candidates, non-native English CV phrasing) rather than an explicit rule, and most agencies never audit their reject pile to catch it.
What AI tools do recruitment agencies use to screen CVs?
The most common are Bullhorn, Manatal, and Textkernel for parsing and matching, Paradox's Olivia for conversational screening, and HireVue for video interview scoring, often layered on top of whichever ATS the agency already had before AI features arrived.
Will AI screening replace recruitment consultants?
Not for the client-facing, judgement-heavy part of the job, but it has already removed a large chunk of the manual CV-reading work, which means agencies now need fewer junior researchers and more consultants who can build client relationships and close placements.
How much does it cost a small recruitment agency to add AI screening?
Bolting a screening layer onto an existing ATS typically runs from a few hundred pounds a month for a smaller off-the-shelf tool up to several thousand for a custom-built pipeline with proper bias auditing, with most small agencies landing somewhere in the low four figures monthly once you include the setup work.
