When should a startup hire its first recruiter?
Hire when your open roles pass what one person can carry, not when you reach a headcount. The benchmark for a manageable load is about 13 open roles at a time. AI companies of 11 to 50 people are hiring their first recruiter into a median of 23. Most founders wait until the job is already impossible.
What founders actually do
I read the job boards of 192 AI companies and counted every live role. Here is the part that matters: 31% of them were hiring a talent person at that moment, 58 companies out of the 190 with a readable board. So this is not a rare decision. It is one most founders in this market face within a year or two.
What they are hiring into is more interesting than when.
| Company size | Companies | Median open roles | Talent people being hired |
|---|---|---|---|
| 1 to 10 | 9 | 22 | 1 |
| 11 to 50 | 19 | 23 | 1 |
| 51 to 100 | 6 | 32 | 1.5 |
| 101 to 250 | 7 | 88 | 2 |
Look at the first two rows. A company of eight people and a company of forty are both hiring their first recruiter into roughly the same number of open roles, 22 and 23. Headcount is not the trigger. Open roles are.
What one recruiter can carry
Between 20 and 30 open roles is the point the benchmarks describe as the breaking point: intake meetings get skipped, active sourcing stops, and the recruiter turns into someone who processes inbound applications.
On output, Ashby’s 2026 data puts a recruiter at 5.0 business-role hires a quarter, so about 20 commercial hires a year. Blended across all role types it is nearer 29, and technical hiring is slower.
The gap, and it is the whole point
A manageable load is about 13 open roles. The AI companies in this census are hiring one recruiter into 23. They are hiring a person into a job the benchmarks already classify as beyond one person, on the day it starts.
That is not an argument against hiring a recruiter. It is an argument against waiting until the number is 23 before you start thinking about it, and against assuming one hire fixes it.
So what should you actually do
| Option | What it costs | How fast | When it is right |
|---|---|---|---|
| Hire in-house | About $215,000 a year, loaded | Two to three months to hire, then ramp | You have steady hiring at a volume that fills their year, and you can wait |
| Use an agency | About 20% of base, per hire | Days to start | You have one or two roles and no intention of hiring continuously |
| Embedded | A monthly fee, whatever the plan turns out to be | Starts now | You have several roles, no recruiter, and a founder who has become the bottleneck |
The honest version: if you have one or two roles, use an agency and stop reading. If you have steady hiring that will fill someone’s whole year and you can wait three months, hire in-house. Embedded is for the space in between, which in this market is most companies most of the time.
Questions founders ask
At what headcount should we hire a recruiter?
Headcount is the wrong trigger. In this census, companies of 1 to 10 people were hiring their first recruiter into a median of 22 open roles, and companies of 11 to 50 into 23. The number of open roles barely moved, so the size of the company is not what creates the need. The open roles are.
How many roles can one recruiter actually fill in a year?
About 20 for commercial roles, using Ashby’s figure of 5.0 business-role hires per recruiter per quarter. Blended across all role types the figure is closer to 29, and for technical roles it is lower. Twenty is the honest number to plan with for sales, marketing, product and delivery hiring.
How many open roles can one recruiter hold at once?
Gem’s 2026 benchmark puts the average at 13.4 open roles. Between 20 and 30 is where intake meetings get skipped and active sourcing stops, and the recruiter becomes someone who processes inbound applications instead.
What does an in-house recruiter actually cost?
Around $215,000 a year once salary, employer costs, tooling and a seat are included. It also takes two to three months to hire one, and then they ramp. The cost does not stop when your hiring does.
Is an agency cheaper than hiring someone?
For one or two hires, yes. At 20% of a $180,000 base that is about $36,000 a hire, so the cost scales with your plan forever. Agencies are the only option whose cost grows every time you succeed.
What is embedded hiring?
Someone works as your hiring function for a monthly fee, inside your tools and your process, rather than sending CVs from outside. The fee does not change with the number of hires, and it stops when you stop.
When is embedded the wrong answer?
Below about three hires. If you have one or two roles, a single search on a percentage fee is cheaper and simpler, and saying so is the only reason to believe the rest.
Where these numbers come from
- The census: 192 live AI company job boards read in full, covering 7,053 open roles, across Ashby, Greenhouse, Lever and Workable. 190 had a readable board. September 2026, my own data.
- Open roles per recruiter: Gem, 2026 Recruiting Benchmarks Report.
- Hires per recruiter: Ashby Recruiter Productivity report, May 2026. Business roles 5.0 a quarter, technical 3.8.
- In-house cost: salary plus employer costs, tooling and a seat, for a mid-level in-house recruiter in the US market.
- No source anywhere separates outbound sourcing from processing inbound applications, so I do not claim a difference between them.
I am Ben Jackson. I have recruited commercial teams into AI infrastructure and tooling companies for 17 years, and I publish a live map of 1,000+ of them and the numbers behind it. If you are working out how to hire sales, marketing, product or delivery people into an AI company, here is how I work.