Why Retained Search Survives AI When Contingency May Not
If sourcing is the thing being automated, the model that never sold sourcing is the model that survives.
Contingency recruiting sells finding people. Retained search sells judgement, process and accountability, and charges for the work rather than the outcome. As sourcing becomes something software does, the first model loses the thing it was charging for and the second does not. That is an uncomfortable conclusion for the larger half of the industry and it follows directly from what each model was actually selling.
Last reviewed August 2026. Model analysis drawn from Sortinghat delivery observations.
Key takeaways
- Contingency charged for access. Retained charges for process. Only one of those is being automated.
- Retained fees are paid for work, not for a result. Which makes them robust to the sourcing cost falling.
- The retained failure modes are human. Assessment, references, offer management and counter-offers, none of which automate.
- This does not mean contingency dies. It means the contingency bet moves from speed to accuracy.
What each model was actually selling
Contingency is a bet on being first. You invest effort with no guarantee of payment, wagering that you submit a placeable candidate before three other agencies. Historically that bet paid because access and speed were scarce.
Retained search is a different transaction entirely. The client pays for a defined piece of work: mapping a market, approaching people who are not looking, assessing them properly, and managing a long process to a conclusion. The candidate list is an output rather than the product.
When the cost of producing a candidate list collapses, one of those propositions loses its foundation and the other does not. The full argument on the contingency side is in does contingency still work.
Why the retained work does not automate
Look at what a retained search actually consists of.
Market mapping. Partly automatable and genuinely improved by better search.
The approach. A senior person contacted like a volume prospect concludes the role is not serious. This is the opposite of automatable.
Assessment. Structured, evidence-based and requiring domain judgement about whether somebody can operate at a level.
References and offer management. Entirely relationship work, and where these searches actually fail.
One of four improves with automation. The other three are the fee.
What this means for firms running both
Most firms run a mixed book, and the honest consequence is that the two halves are moving in different directions.
The contingency half faces compression, because everyone is fast now and speed was the differentiator. The response is accuracy, which is a screening standard rather than an effort increase.
The retained half faces a different problem: clients asking why a retainer is necessary if software finds people. The answer is that the retainer never bought the finding, and firms that cannot articulate that will lose the argument.
Moving work from one model to the other
The practical opportunity is that several role types currently worked on contingency should not be.
Genuinely scarce skills. A forty-person national population contested by several agencies is not a rational contingency bet, as covered in genuinely scarce roles.
Confidential mandates. Where the search cannot be advertised and the population is narrow by design.
Roles where the client keeps changing the brief. Retained terms fund the discovery work that changes it.
The conversation is not about the fee. It is that the work cannot be done properly under contingency risk, which is a more persuasive argument than a capability pitch.
Where this argument overreaches
Two honest qualifications, because the case above can be pushed too far.
Retained is not defensible on commodity roles. A client will not pay a retainer for a role five agencies can fill in a week, and asking damages the relationship.
Contingency volume still funds many firms. Declaring it obsolete is easy for a commentator and expensive for an operator. The model changes; it does not disappear.
The useful conclusion is narrower: the share of your book that should be retained is larger than it currently is, and the reason is structural rather than commercial ambition.
Frequently asked questions
Will AI replace executive search?
It improves market mapping and research, which is one of four components. The approach, the assessment and the offer and reference stages are relationship and judgement work, and they are the majority of what a retained fee actually buys.
Why is retained search more resilient than contingency?
Because contingency historically charged for access and speed, both of which are being automated. Retained search charges for a defined piece of work whose most valuable parts are human, so falling sourcing costs do not undermine it.
Which roles should move from contingency to retained?
Genuinely scarce skills where the national population is small and contested, confidential mandates that cannot be advertised, and roles where the client keeps changing the brief because discovery work needs funding.
How do you justify a retainer when clients can search themselves?
By being clear that the retainer never bought the finding. It buys mapping, a credible approach to people who are not looking, structured assessment, and management of a long process through references and offer.
Is contingency recruiting dying?
No, but the differentiator moves. Speed was the historic edge and it is commoditising, which shifts the bet towards accuracy. Firms that verify properly rather than simply moving fast will do better than those that do not.
Can retained terms be used for ordinary roles?
Not credibly. A client will not pay a retainer for a role several agencies could fill in a week, and asking damages the relationship. The argument works where the work genuinely cannot be done under contingency risk.
The split worth examining in your own book
Take last year's placements and mark which were genuinely scarce, confidential, or subject to a brief that changed materially during the search. Then check how many of those were worked on contingency.
That overlap is the part of your book being run on the wrong commercial structure, and it is usually where your least profitable months came from.
See a long search stay documented
Bring a retained mandate and we will show you how months of intermittent conversation stay on one record.
Book a demoFounder of Sortinghat, an AI-native ATS and CRM for staffing, search and RPO firms. Writes about recruiter capacity, sourcing economics and what actually changes when AI reaches a delivery desk. More about the author
Related reading