Vendors in the agentic category sell time savings, which is the right claim and an unverifiable one in a demo. This article does the arithmetic transparently so you can substitute your own numbers, see which assumption the answer actually depends on, and know what to measure in a pilot.
All figures below are estimates and clearly labelled as such. Software costs come from our pricing benchmark, where the basis for each vendor estimate is documented; hours are practitioner-reported ranges, not measurements from a controlled study.
The assumptions, stated
| Input | Value used | Basis |
|---|---|---|
| Recruiter fully loaded cost | $115K/year → ~$55/hour | Salary plus ~30% benefits and overhead, US mid-market |
| Manual sourcing hours per role | 15–25 | Practitioner-reported: search, screen, write, follow up |
| Hours per role with an agent | 3–6 | Briefing, calibration grading, shortlist review |
| Roles sourced per year | 20 | Mid-market in-house team baseline |
| Agent subscription | $1K–$5K/year | Our estimate, e.g. Noon |
| Database seat (comparison) | $8K/year | Our estimate, mid-tier hireEZ/SeekOut range |
Two caveats before the arithmetic. The agent hours assume shortlists good enough that the recruiter isn't redoing the screening — the single most important assumption in the model. And the fully loaded rate is what the business pays, not what the recruiter earns; using salary alone understates savings by roughly a third.
The base case: 20 roles a year
| Line | Database + recruiter | Agent + recruiter review |
|---|---|---|
| Software | $8,000 | $1,000–$5,000 |
| Sourcing hours (20 roles) | 300–500 | 60–120 |
| Labour at $55/hr | $16,500–$27,500 | $3,300–$6,600 |
| Total annual | $24,500–$35,500 | $4,300–$11,600 |
| Hours freed | — | ~180–440 |
| Value of freed hours | — | $9,900–$24,200 |
The headline is not the software saving; it's the 180–440 recruiter hours, which is between one and two-and-a-half months of a full-time recruiter's year. What that is worth depends entirely on what the hours get redeployed into — more roles worked, faster interview loops, better candidate experience, or nothing at all.
Break-even, which is lower than people assume
Turn the model around: how much time must an agent save to justify itself?
| Subscription | Hours to break even (at $55/hr) | Per role, 20 roles/year |
|---|---|---|
| $1,000 | 18 hours/year | ~0.9 hours |
| $2,500 | 45 hours/year | ~2.3 hours |
| $5,000 | 91 hours/year | ~4.5 hours |
| $15,000 | 273 hours/year | ~13.6 hours |
At the low end of the market, break-even is about one hour saved per role — a bar almost any functioning automation clears. At $15K it needs to remove nearly the whole manual sourcing workload, which is a much stronger claim and should be tested hard in a pilot. The practical implication: for cheap agents, the licence decision is nearly free and the real question is whether shortlist quality holds. For expensive ones, the vendor is making a claim about replacing the entire motion.
Sensitivity: which assumption decides the answer
Vary one input at a time from the base case (20 roles, $2.5K subscription, 20 manual hours, 4 agent hours):
| Change | Effect on annual value of freed hours | Verdict |
|---|---|---|
| Recruiter cost $40/hr instead of $55 | $17,600 → $12,800 | Still strongly positive |
| Only 8 roles a year | $17,600 → $7,040 | Positive, less compelling |
| Manual sourcing only 10 hours/role | $17,600 → $6,600 | Positive; you were already efficient |
| Agent needs 10 hours/role (weak shortlists) | $17,600 → $11,000 | Still positive, but check quality |
| Agent needs 18 hours/role (poor calibration) | $17,600 → $2,200 | Case collapses |
| Reply rate halves versus manual outreach | Fewer candidates per role; may need 2× volume | Depends on brand cost |
Notice which inputs matter. Recruiter cost and role count move the answer proportionally but don't change the sign. The variable that can flip the decision is agent hours per role — i.e. shortlist quality — and secondarily reply rate. Everything else is noise by comparison.
That is exactly why the calibration loop is the thing to interrogate in evaluation, and why a pilot should measure recruiter hours including rework. A tool that surfaces 40 candidates a week and requires a recruiter to re-screen all of them has not automated screening; it has moved it.
Four conditions where the case collapses
1. Shortlist quality is poor. Covered above: rework restores the hours you were buying back. This is the dominant risk and the only one a pilot measures directly.
2. No capacity to absorb the pipeline. If hiring managers can't interview more people, extra qualified candidates queue up and expire. The constraint has moved downstream and the ROI is theoretical. Check interview capacity before buying throughput.
3. Volume is unpredictable on per-role pricing. Outcome-based pricing converts a fixed cost into a variable one, which is usually good — until a hiring surge produces an unbudgeted bill. Ask for a cap or committed-volume pricing.
4. Briefs can't be specified. Agents execute a brief faithfully. Where the requirements are genuinely undiscovered — a first executive hire, a function new to the company — the work is definition, and a good human sourcer earns their cost by arguing with the hiring manager. See sourcing hard-to-fill roles.
For agencies the model changes shape: freed hours convert into placement fees rather than relief, which raises the upside considerably — but only if client demand exists to absorb the capacity. The agency-specific version is in our agency stack guide.
Against the alternatives
| Option | Est. annual cost, 20 roles | What you're really buying |
|---|---|---|
| Agent plus recruiter review | $4K–$12K | Screening and outreach executed; you keep judgement |
| Database plus recruiter | $25K–$36K | Control and depth; you keep the hours too |
| Dedicated sourcer (fully loaded) | $80K–$120K | Craft judgement on ambiguous briefs |
| Agency, 4 placements at 18% of $120K | $86K | Outsourced risk and speed on specific roles |
Read this as an order-of-magnitude comparison, not a like-for-like one: an agency carries risk you don't, and a dedicated sourcer does things an agent can't. But the spread explains why the agentic category has moved fast — it competes with labour costs, not with software budgets.
To build your own version of this model, you need two numbers most teams don't have: your actual sourcing hours per role and your qualified-candidates-per-week baseline. Both take two weeks of light logging to establish, and without them any vendor's ROI claim is unfalsifiable. The instrumentation steps are in our metrics guide, and the pilot design that tests the assumptions is in our buyer's framework. Tool candidates are in the AI sourcing ranking.