KITHResearch briefing / 01 / November 2026

The independent recruitment desk

GOING
ALONE.

Why the next recruitment business may be built around one good biller, with an operating partner behind them.

The question is no longer whether you need an agency.

It is which parts of one you still need to own.

A different way to organise the work is emerging: own the relationships, the judgement and the fee. Buy the research, systems and delivery capacity you need without recreating the whole firm.

An argument for UK permanent and search recruiters. Economics are illustrative, not a forecast. Source and assumptions on page 6.

Kith / Research01 / 08
KITHGoing alone / The bundle

01 / The structural change

An agency is a bundle.
The desk isn't.

Traditional agencies combine things that require a recruiter with things that merely support one. Going independent used to mean rebuilding both.

Own the advantage
Client relationshipsMandates & feesCandidate trustJudgement & close

These compound with the individual. They are the reason clients come back, and the hardest parts to purchase.

Buy the capacity
Market mappingCandidate researchSearch administrationSystems & support

These can increasingly be provided as software and managed operations. You still specify the outcome and inspect the work.

Retain control
Data & permissionsQuality standardsComplianceContinuity

Delegating the task is not the same as transferring accountability. Contracts, records and exit rights still matter.

For context, one UK sample of 84 agencies reported wages at 67% of net fee income, overhead at 25.4% and EBITDA at 7.5% in H1 2026. These are firm-level expenses, not the share an employee gets. [2]

The implication

The smallest viable recruitment business may no longer need to look like a small recruitment agency.

Structure / not hype02 / 08
KITHGoing alone / The software reset

02 / The software reset

Every ATS sucks.
That isn't the point.

The old problem was choosing the right software. The new problem is getting an entire desk to work the way you do.

Making software is cheaper.

Code generation is becoming ordinary work. In Stack Overflow's 2026 survey, 69% of respondents used AI to generate code in familiar areas. Only 9.6% used an AI answer as-is. Build speed is not the same as operational reliability. [6]

The value is moving elsewhere.

Databases and generic screens are easier to reproduce. Yet only 18.6% in Stack Overflow's 2026 developer survey said they had built, or planned to build, tools to replace paid vendors. Over half of that group used AI. The shift is real, but SaaS has not disappeared. [6]

Capital markets as a directional signal

6.9×→4.2×
20212025

Median SaaS M&A enterprise value / trailing revenue, Software Equity Group. Deal multiples reflect rates, growth and buyer mix as well as product defensibility. They do not measure the price of an ATS, nor prove AI caused the reset. Buyers continue to reward products deeply embedded in workflows. [7]

Software isn't worthless.
Undifferentiated software is becoming less scarce.Kith's editorial view. The question is no longer what software you can buy, but what work you can stop doing.
Value / not valuation advice03 / 08
KITHGoing alone / The execution gap

03 / The execution gap

It's never been easier
to automate the
wrong thing.

Build enough Lego blocks and you get a pile of Lego. An operating system only exists when the pieces fit, the handoffs work, and somebody owns the outcomes.

42%

of surveyed companies abandoned most of their AI initiatives before production, up from 17% the previous year. S&P Global, across industries. This is not a recruitment-specific failure rate. [8]

10%

of staffing firms surveyed by Bullhorn in 2026 said agentic AI was embedded across their entire workflow. The report identifies data readiness, security and implementation plans as obstacles. Vendor survey. [9]

Redesign beats bolt-on.

In McKinsey's 2025 AI survey, high performers were nearly three times as likely as others to report fundamentally redesigning workflows. An association, not proof that redesign alone caused better results. [10]

Before you automate, ask

01What does the recruiter never need to touch again?
02Where does a human still need to make the call?
03Who spots a bad outcome and changes the system?
Workflows / not tools04 / 08
KITHGoing alone / The arithmetic

04 / The economics

The upside is real.
So is the exposure.

A useful comparison starts with fees actually collected, not a fantasy about saving an agency's overhead. The illustration below is deliberately simple. Change the inputs before you act.

Illustrative annual net fee income

£300,000

An established recruiter's prior run-rate, not a prediction for year one.

The employed alternative

£75,000

Illustrative salary at 25% of fees. Not a market average. Owner earnings and salary are not like-for-like.

Fees collected

100%
£300k

Build-and-run delivery allowance

Sourcing, human review, tools and workflow support

15%
−£45k

Other running costs

Tools, professional services, insurance and admin

Assumed
−£24k

Available to the owner

Before owner pay, taxes, pension and reserves

77%
£231k

What if collections fall?

100% fees
£231k
75% fees
£167k
50% fees
£104k

All figures are Kith's illustrative arithmetic: collected fees × 85% less £24,000 fixed cost, rounded to the nearest £1,000. They omit setup, working-capital needs, VAT, bad debt, fee rebates, benefits and any changes to the time needed to win work. The 15% is an illustrative revenue-share assumption, not a supplier quotation. The 25% employed share is also assumed; gross pay and pre-tax owner cash are not directly equivalent.

Assumptions / not projections05 / 08
KITHGoing alone / The operating partner

05 / A different partnership

Don't buy tools.
Build a desk
that runs.

The recruiter should not be the systems integrator. The relationship needs to sit inside the daily workflow, with someone responsible for building it, operating it and improving it.

01

Design together

Start with the actual brief, buyer, candidate market and handoffs. Agree what a useful shortlist looks like before automating anything.

02

Assemble the blocks

Research agents, an ATS if needed, evidence checks, candidate pages, screening, client updates and billing. Use interchangeable components rather than buying a fixed workflow.

03

Run the operation

Software handles repeatable work. A delivery team manages exceptions and quality. The recruiter owns relationships, candidate conversations, judgment and closing.

04

Learn from each search

Every rejection, changed brief and candidate conversation improves the next pass. The supplier stays accountable for uptime, provenance, handoffs and data portability.

The point is to pick up the phone.
Not become your own IT department.
A partner / not a subscription06 / 08
KITHGoing alone / The risk

06 / The case against

Not everyone
should go alone.

Lower fixed costs do not create clients. Technology does not make a bad desk commercially viable.

01

Where do the first three mandates come from?

Contacts are not contracted work. If you cannot name credible first clients and route to market, solve that before buying infrastructure.

02

Can you fund the first unpaid quarter?

Build a monthly cash-flow plan covering the sales ramp, fees outstanding, replacement guarantees and any client concentration. Billing is not collection.

03

Can you leave cleanly?

Review restrictive covenants, confidentiality, intellectual property and candidate data. Do not transfer your employer's records. Take legal advice on your actual contract.

04

Who owns the compliance decision?

Check Conduct Regulations obligations, fair selection, privacy notices, retention and the role of any AI supplier. Outsourcing execution does not outsource responsibility. [3–5]

05

What happens if your provider fails?

Insist on data export, service levels, evidence standards and a workable fallback. A desk that cannot operate without one supplier has merely traded dependencies.

If you have no repeatable way to win mandates, you don't yet have an independent desk. You have an operating stack.
Counterargument / decision gates07 / 08
KITHGoing alone / Conclusion & method

07 / The thesis

The agency
of one.

Keep the client.
Keep the call.
Change the infrastructure.

A viable independent recruiter still needs a way to win mandates, place people and get paid. They may no longer need to hire a research team or assemble a fragmented software stack to do it. The opportunity is not a cheaper ATS. It is a smaller business with an operating partner behind the biller, built around where a recruiter actually creates value.

Evidence and method

  1. [1] Bullhorn, GRID 2025. 14.6 hours searching: weighted estimate, global vendor survey. Source.
  2. [2] Recruitment Accountants, Q2 2026. Dataset of 84 UK agencies. 67% wages, 25.4% overhead, 7.5% EBITDA as shares of NFI, not recruiter pay. Source.
  3. [3] ICO, Recruitment and selection. UK data protection guidance. Source.
  4. [4] GOV.UK, Responsible AI in Recruitment. UK guidance on oversight, fairness and assurance. Source.
  5. [5] GOV.UK, Conduct Regulations 2003. GB agency guidance; obligations vary by activity and jurisdiction. Source.
  6. [6] Stack Overflow, 2026 survey. Developer AI usage and build-versus-buy responses; self-selected survey. AI data; vendor replacement. Data © Stack Exchange Inc., ODbL 1.0.
  7. [7] Software Equity Group, 2026 Annual SaaS Report. Median SaaS M&A EV/TTM revenue: 6.9x in 2021 and 4.2x in 2025; no inference about product prices. Source.
  8. [8] S&P Global, October 2025. Enterprise survey on abandoned AI projects, all industries; not staffing. Source.
  9. [9] Bullhorn, GRID 2026. Nearly 2,300 recruitment respondents, vendor survey; 10% embedded agentic AI end to end. Source.
  10. [10] McKinsey, State of AI 2025. High performers nearly three times as likely to redesign workflows; correlation only. Source.
  11. Economics, page 5. Original illustrations only; 15% revenue share and £24k costs are assumptions, not a Kith offer or industry forecast. No tax, pension, setup costs or working capital modelled.

About this briefing. Published by Kith, which builds recruitment technology and provides human-in-the-loop delivery for recruiters. Kith has a commercial interest in the operating model discussed. Research and editorial judgement should be assessed accordingly.

Further reading. A commercial example of the model is described at kith.contact. Version 3, November 2026. UK permanent and search recruitment focus.