How to Build an In-House Recruiting Function That Doesn't Depend on Headhunters

You can replace most headhunter spend with an in-house function, but not all of it and not in a quarter. A company making ten hires a year at 70,000 euros average salary pays 140,000 to 210,000 euros in contingency fees at the standard 20 to 30 percent. That money buys two things: sourcing capacity and evaluation judgment. Both can be built internally at that volume, in this order.

The arithmetic that makes this worth doing

Contingency fees are quoted at 20 to 30 percent of first-year salary across most of the European market, and retained search runs higher. That range is the one constant in every agency contract I have signed or issued. Here is the cost at a normal mid-market salary.

Hires per year	Average salary	Fees at 20%	Fees at 30%
5	70,000 EUR	70,000 EUR	105,000 EUR
10	70,000 EUR	140,000 EUR	210,000 EUR
20	70,000 EUR	280,000 EUR	420,000 EUR

Now the honest side, because in-house is not free. Against those fees you have to set a recruiter's fully loaded cost at your own market rate, a recruiting system, job advertising, a sourcing licence or two, and the hiring-manager hours an agency currently absorbs for you. SHRM's widely cited benchmark puts average cost per hire at roughly $4,700 before any agency fee.

At five hires a year the case is weak. At ten it is strong. At twenty it is not really a debate. The break-even is not a number of hires, it is a level of process maturity: an internal recruiter with a bad process is more expensive than an agency, because you pay the salary and still pay the fees on the roles they cannot fill.

One thing the fee buys gets forgotten in every business case: most contingency agreements carry a replacement guarantee, so if the hire leaves inside three to six months you get a rebate or a free re-run. In-house that risk is yours, and the widely cited figure for a failed hire is around 30 percent of first-year salary. What a hire really costs is in the true cost of hiring.

Where an agency is still the right call

I ran an agency, and I will say plainly which mandates were worth the fee and which were us being paid for work the client could have done itself.

Keep it in-house	Use an agency
Roles you hire repeatedly (sales, support, engineering levels you know)	Genuinely scarce senior specialists where the qualified pool is small and passive
Roles where your employer brand does the work	Confidential replacements you cannot advertise
Anything where you already have rejected finalists on file	First hire in a country where you have no network at all
Junior and mid-level volume hiring	Sudden volume spikes your team has no capacity to absorb

The left column is a process problem. The right column is a network or discretion problem, and networks take years to build. That is the strategy: bring the process problems in, keep buying the network problems.

Stage 1: Define the role before anyone sources

This is where most in-house functions fail, and it fails quietly. A job description is a marketing document; a brief is a decision document. They are not the same thing.

A usable brief names the three to five things the person must be able to do in their first year, the evidence that would prove each one, and what you are willing to trade away. Run a 45-minute intake and force the ranking. If the hiring manager says all eight requirements are must-haves, they have not decided yet, and you will spend six weeks sourcing against a wish list.

Agencies are good at this because a bad brief costs them money directly. Internally, nobody feels that cost until the fourth rejected shortlist.

Stage 2: Make the candidates you already have searchable

If you have been hiring for a few years, everyone who ever applied is still in your system, treated as an archive. Two groups there are worth real money:

• Silver medallists. Strong candidates rejected only because someone else was marginally stronger. Nothing about them changed when they lost.
• Right person, wrong time. People who applied when you had no budget, no headcount, or no matching role.

Nobody mines this because most systems only search keywords. A search for "Kubernetes" misses the candidate whose CV says "container orchestration on GKE". You conclude the archive is empty when it is only unreadable. Scoring on demonstrated and adjacent skills turns that archive into a pipeline; the practical build is in how to build a talent pool.

Cost per hire from your own pool is close to zero, and almost nobody does this work first.

Stage 3: Make outreach systematic rather than heroic

The agency advantage in sourcing is not mysterious. Outreach is somebody's entire job, done to a cadence, tracked. Internal outreach dies because it is the fifth priority of someone who also runs interviews, scheduling and offers.

What makes it survive:

1. A fixed block in the week that does not move. Two hours, twice a week, sourcing only.
2. Three to four touches per candidate, not one message and hope.
3. A first message that references the person's actual work, not their job title.
4. Response rates tracked per role, so you learn in week two that the message is wrong, not in month two.

A personalised sequence and a generic template are the difference between a pipeline and silence. Use the template as a skeleton, then rewrite it for the person.

Stage 4: Structure the interview so it produces comparable evidence

Unstructured interviews produce confident opinions that cannot be compared across candidates or interviewers. The fix is unglamorous: the same core questions for every candidate, defined criteria, scores recorded before the panel talks, and written evidence attached to each criterion rather than a bare number.

The blocker is never the theory, it is that hiring managers do not fill in scorecards. Two fixable reasons: they are handed an empty form three days later, when the detail has evaporated; and many systems charge per seat, so finance caps the licence count and feedback ends up in messages that never reach the record.

Stage 5: Measure outcomes, or the function never improves

An agency rarely finds out whether the people it placed worked out: it has no visibility into your first-year reviews and little reason to look. That feedback loop is the biggest structural advantage of doing it yourself, and most internal teams throw it away by measuring only time to hire.

Track, per hire: 12-month retention, hiring-manager satisfaction at six months, performance rating at the first review, offer acceptance rate, and stage-by-stage time. Then look at which sources, interviewers and criteria actually predicted the people who succeeded, and change the brief accordingly.

What AI removes, and what it does not

AI is why this playbook is realistic for a small team in 2026 and was not in 2018. It removes the research layer an agency used to sell you: screening every application against the role's real requirements, keeping a large archive searchable by evidence, and drafting outreach.

Pickr is the AI-native recruiting platform that gives an internal team the screening and evaluation capacity agencies used to sell as a service: every candidate scored continuously on evidence of skills, including adjacent and transferable ones, interviews transcribed into scorecards pre-filled with evidence mapped to each criterion, and free seats for every interviewer and hiring manager. The same system runs multi-client pipelines for agencies, which is the point: both sides do the same work. Candidate data is hosted in Frankfurt, Germany, GDPR compliant, with a data processing agreement included. I built Pickr, so discount this paragraph accordingly; the arithmetic above stands without it.

What AI does not remove: the intake conversation that finds out what the hiring manager actually wants, the judgment call between two strong finalists, the closing conversation when a candidate holds a counteroffer, and accountability. A machine can propose and evidence a recommendation. A person decides and owns it. Which parts of the headhunter's job survive is the argument in what AI actually replaces in headhunting.

Where to start

Do not start by hiring a recruiter. Start by finding out what your current process costs and where it leaks. The free recruiting audit is an eight-question wizard, about two minutes, no signup, that returns what your process costs per month in hours and euros. The deeper version connects read-only to your existing system and reports funnel drop-off stage by stage, which stages run slower than comparable teams, and how often scorecards get completed. The key is never stored, the data stays in the EU, and you can delete it.

Then work the stages in order: brief and candidate archive in month one, outreach cadence and interview structure in month two, your two most repeatable roles in-house in month three, the hard one still with an agency.

The decision is not whether agencies are worth it. For some roles they clearly are. It is whether you are still paying 20 to 30 percent on roles you could fill yourself with a better brief and a searchable archive of the people who already applied, and whether you will spend one quarter finding out.