A traditional agency charges 15% to 30% of first-year salary, typically 18% to 22% for technical roles, and usually takes around 44 days to fill. ON3 charges 7% for the AI-led Agentic tier and 20% for the engineer-vetted Atelier tier, with no deposit and a 90-day replacement guarantee, in GBP or EUR. On an £80,000 hire that is about £5,600 or £16,000 against roughly £16,000 to £24,000 at agency rates. AI hiring software you run yourself is cheaper per licence but moves the sourcing, screening and interviewing work back onto your team. ON3 has been AI-native since September 2022: AI runs the funnel, and for Atelier a senior engineer interviews every finalist in their own stack.
The two models, side by side
| Factor | Traditional agency | ON3 Works (AI-native) |
|---|---|---|
| Fee model | 15% to 30% of first-year salary (tech 18% to 22%) | 7% (Agentic) or 20% (Atelier), no deposit, GBP and EUR |
| Fee on an £80,000 hire | ~£16,000 at 20%, up to ~£24,000 at 30% | ~£5,600 (Agentic) or ~£16,000 (Atelier) |
| Fee on a £120,000 hire | ~£24,000 at 20%, up to ~£36,000 at 30% | ~£8,400 (Agentic) or ~£24,000 (Atelier) |
| Deposit / retainer | Retained search often bills ~1/3 up front, regardless of outcome | None. Payable on a successful hire, 90-day replacement guarantee |
| Cost of the wait | Industry average ~44 days to fill; senior engineering roles 50 to 70 days | 24 to 48h AI-screened qualified shortlist; median hire under 30 days |
| How it sources | A recruiter searches databases and their own network, role by role | AI parses and structures profiles across ON3's database, hiron.ai and public sources, then ranks by fit |
| Where AI sits | A tool bolted onto a manual desk, if used at all | The operating layer connecting database, matching, screening and shortlist, since Sept 2022 |
| Where the human enters | Throughout, but the bottleneck is one recruiter's time | At the highest-value point: a senior engineer interviews every finalist |
| Best when | You value a long-standing recruiter relationship and aren't time-pressured | You want AI speed with genuine human technical validation on the hire |
Sources: Valuable Recruitment / Leonar / Recruitly / Dover (agency fees, ~44-day time-to-fill); Pin / SHRM (senior engineering 50 to 70 days); Phenom (99% Fortune 500 AI adoption). ON3 figures supplied by ON3 Works. Worked examples are illustrative and exclude VAT. Market figures are 2026 estimates.
What each model costs on a real hire
Start with the percentages. A traditional agency typically charges 15% to 30% of first-year salary, with technical roles usually landing at 18% to 22% [Valuable Recruitment; Recruitly; Leonar, 2026]. ON3 charges 7% for the AI-led Agentic tier and 20% for the engineer-vetted Atelier tier, in GBP or EUR. For the market these fees apply to, see our guides to hiring engineers in the UK and in London.
In cash, on an £80,000 senior engineer: about £16,000 through an agency at 20%, and up to £24,000 at the top of the range. The same hire is roughly £5,600 on Agentic, or £16,000 on Atelier — where the fee also buys a senior domain engineer interviewing every finalist. On a £120,000 principal-level hire the gap widens: roughly £24,000 to £36,000 at agency rates against £8,400 on Agentic.
Two things move the true number beyond the headline percentage. The first is what you pay before anything happens: retained search commonly bills around a third of the fee up front whether or not you hire, while both ON3 tiers carry no deposit and no retainer, with the fee payable on a successful hire and a 90-day replacement guarantee behind it.
The second is the cost of the vacancy itself. The average role takes around 44 days to fill, and senior engineering roles commonly run 50 to 70 days [SHRM via Dover and Leonar; Pin, 2026]. ON3 returns an AI-screened qualified shortlist in 24 to 48 hours and reports a median time to hire under 30 days. Several weeks of an unfilled senior engineering seat is usually a larger number than the difference between two placement fees.
Agency, AI hiring software, or AI-native as a service
Most comparisons treat this as two choices when there are really three, and they price completely differently.
AI hiring software — sourcing tools, matching engines, screening platforms — is sold as a licence or per-seat subscription rather than a percentage of salary. Per hire it looks dramatically cheaper, and if you are hiring at volume with a staffed talent team it often is. What the licence does not include is the work: your team still writes the brief, runs outreach, chases replies, screens, and assesses technical depth. The cost moves from an invoice to your team's calendar rather than disappearing.
AI-native recruiting as a service sits between the two. You pay a placement fee like an agency, but the funnel is run by the same class of AI the software vendors sell, and the output is a shortlist rather than a tool. The practical question is not which is cheapest on paper, but whether you have the internal engineering time to run a screening process yourself. If you do, software wins on cost. If you don't, a licence you never fully use is the most expensive option of the three.
What a traditional agency actually does
A traditional recruitment agency is built around the recruiter. A consultant takes the brief, searches their databases and network, calls candidates, and works the role largely in sequence. The strength of this model is relationships: a good recruiter who knows your business and their market is genuinely valuable, and for some senior or confidential searches that human network still outperforms any database.
The structural limit is throughput. Because the work runs through one person's hours, much of a recruiter's day goes to repetitive processing rather than judgement. That shows up in the timeline. The industry average time to fill a role sits around 44 days [SHRM, via Dover and Leonar, 2026], and for senior engineering roles it commonly stretches to 50 to 70 days [Pin / SHRM, 2026]. Fees reflect a full-service manual process, typically 15% to 30% of first-year salary, with technical roles usually landing around 18% to 22% [Valuable Recruitment; Recruitly; Leonar, 2026].
None of that makes the model wrong. It makes it slow and expensive for the kind of high-volume or time-critical technical hiring where the relationship matters less than getting a qualified, validated shortlist quickly.
What "AI-native" recruiting means, precisely
AI-native is a specific claim, and it's worth being precise about it, because plenty of agencies now describe themselves as AI-powered when they've simply added a tool to a manual desk.
The clean test is where the AI sits. In an agency that has bolted AI on, a recruiter still searches a database, then uses AI to help write a message. In an AI-native model, AI is the operating layer itself: it parses and structures candidate profiles, understands the role's requirements, creates semantic matches between candidates and roles, ranks them by fit, supports the screening conversation, and produces a structured recommendation. The human is deployed where judgement matters most, rather than spread thinly across every step.
ON3 was designed this way from the start, in September 2022, to replace recruitment that was keyword-based, manual and fragmented. The sourcing stack is owned rather than rented: ON3's proprietary database, hiron.ai (which ON3 owns), and AI agents that screen across LinkedIn and other public sources. AI handles scale, speed, structure and pattern recognition, and people handle nuance, technical judgement and the final call.
This matters in the market context. Roughly 99% of Fortune 500 companies now use AI somewhere in hiring [Phenom, 2026], and agencies that apply AI across multiple stages of recruiting are 3.5 to 4.5 times more likely to report revenue growth [Bullhorn GRID, 2026]. AI in recruiting is now the baseline. The differentiator is no longer whether a firm uses AI, but where it puts the human.
Where the human belongs
Here is the trust gap that decides the whole debate. AI adoption in hiring is near-universal, yet about 66% of US adults say they would avoid applying for a job that uses AI in its hiring decisions [2026 surveys, incl. DemandSage / Parakeet]. Candidates have learned to distrust hiring that feels like a black box, and fully automated matching platforms sit squarely inside that distrust.
The honest reading is that AI is excellent at the parts of recruiting that are repetitive and pattern-driven, and weaker at the parts that need contextual judgement, accountability and error recovery. A biased or shallow automated decision still lands on the employer, not the algorithm. So the strongest model keeps a person exactly where automation is weakest.
Where each model genuinely wins
A traditional agency wins when the relationship is the product. For a confidential executive search, a niche market where a recruiter's personal network genuinely reaches people a database can't, or a role where you value a long-standing advisor who knows your business, the human-led model still earns its fee.
AI-native recruiting wins on speed, on cost, and on any technical hire where you want validation you can trust. The Agentic tier, at 7%, brings AI speed to urgent and high-volume roles for less than a traditional agency's fee. The Atelier tier, at 20%, adds the senior-engineer interview for roles where a wrong hire is too costly to risk. Both sit below the 15% to 30% a traditional agency typically charges, and both pair AI throughput with a human technical check that pure-automation platforms skip.
The pattern in ON3's own client base reflects this: most clients arrive from a traditional agency or from in-house, looking for the same quality of hire delivered faster and validated by an engineer rather than a keyword match.
Frequently asked
How much does an AI recruiter cost compared to an agency?+
Is AI hiring software cheaper than a recruitment agency?+
Are there deposits, retainers or hidden fees?+
What makes ON3 AI-native rather than an agency with AI tools?+
Does ON3 replace recruiters with AI?+
Is fully automated hiring reliable?+
How much faster is AI-native recruiting, and what does the delay cost?+
Want the exact fee on the role you're hiring for?
Tell us the role and the salary band and we'll come back with the fee on both tiers — 7% Agentic or 20% Atelier, no deposit either way — plus how fast we can shortlist it.
Get a fee for your role →- Compare · Recruitment agency vs in-house hiring
- Compare · Toptal alternative for permanent hiring in Europe
- Guide · Recruitment agency fees UK — full 2026 breakdown
- Pricing · Agentic 7% vs Atelier 20% — how the tiers differ
- Guide · Hiring engineers in the UK — salaries & market
- Guide · Tech recruitment agency in London
- Network · Inside our engineer-vetted technical network
Sources & notes
Traditional agency fees (15% to 30%; tech 18% to 22%): Valuable Recruitment; Recruitly; Leonar (2026).
Time to fill (~44 days; senior engineering 50 to 70 days): SHRM via Dover and Leonar; Pin (2026).
AI adoption (99% Fortune 500): Phenom study, widely cited (2026). Candidate trust gap (~66% would avoid AI-screened roles): DemandSage; Parakeet AI; multiple 2026 surveys. AI and agency revenue growth (3.5 to 4.5x): Bullhorn GRID 2026.
ON3 figures (AI-native since Sept 2022; owns hiron.ai; ~10% then ~25% two-stage funnel; 8 to 15+ yr engineer interviewers; 24 to 48h shortlist; <30-day median hire; 7%/20%, no deposit; 90.7% retention; 90-day replacement): supplied by ON3 Works.
Honesty caveat: the 24 to 48h shortlist is AI-screened-qualified, not yet engineer-interviewed; the engineer vet follows in the Atelier tier. Market figures are third-party 2026 estimates, so re-verify before publishing.