We Solved Sourcing. Why Didn't We Solve Hiring?

AI has made one part of recruiting undeniably better: finding people.

 

A recruiter can now take a mandate that once required days of research and identify a highly specific group of candidates in a fraction of the time. Industry, company background, function, skills, education, years of experience, funding stage, geography and other variables can all be layered into a search with remarkable precision.

 

That should have made hiring substantially faster.

 

So far, it hasn't.

 

That is the part I find interesting.

 

I've spent more than 25 years in executive search. For most of that time, one of the practical reasons companies hired recruiters was that searching the market was difficult and time-consuming. Hiring managers had businesses to run. They were not supposed to spend hours digging through job boards, resumes and databases trying to determine who might be capable of doing the job.

 

The recruiter absorbed that burden, but good executive search was never limited to identifying names.

 

A hiring manager explained the business problem. The recruiter learned what mattered, understood the market, challenged assumptions when necessary, identified people who had solved similar problems and narrowed the field to a small number of candidates worth the hiring manager's time.

 

Search required access to the market, but it also required judgment.

 

AI is changing that equation.

 

Sourcing became dramatically more efficient. Hiring didn't.

 

The data is difficult to ignore.

 

Ashby's recruiting data, covering more than 100 million job applications, shows applications per hire rising from roughly 100 in early 2021 to more than 300 during 2025. Recruiters are processing substantially more candidate volume and are also producing more hires.

 

Yet time-to-hire has remained remarkably stable. Business hiring has generally remained in the 36-to-41-day range, while technical hiring has remained roughly 45 to 51 days.

 

We have become much more efficient at finding and processing people without seeing a comparable improvement in how long it takes to make a hire.

 

I think one reason is that we have spent much of the productivity gain on greater specificity.

 

Better search has changed what companies expect to find.

 

Modern talent technology gives companies visibility into the labor market that did not exist even a few years ago. Once nearly everyone becomes searchable, it is understandable that hiring teams begin to believe they should be able to find someone who satisfies almost every preference.

 

A search may begin with legitimate requirements around experience and capability, then add preferred companies, educational background, company stage, years of experience, technical exposure, customer exposure, geography and compensation.

 

None of those criteria is necessarily unreasonable on its own. The problem is what happens when they are treated as cumulative requirements.

 

A search platform may identify 25 people who appear to satisfy the specification, but that does not mean the company has 25 viable candidates. Some will have no interest in leaving. Others will not want the company, the stage, the work model or the location. Some will be outside the compensation range. Others will have alternatives they consider more attractive.

 

The number of people who can be identified is not the same as the number of people who can realistically be recruited.

 

That distinction matters because increasingly sophisticated search tools can create an artificial sense of supply. A hiring team sees a group of highly matched profiles and reasonably concludes that the market contains plenty of candidates. In reality, it may contain plenty of identifiable people while offering very few attainable ones.

 

Company pedigree can make this problem worse.

 

Consider the common desire to recruit from a company such as Palantir.

 

There may be very good reasons to look there. Palantir hires strong people and develops talent in environments that may be relevant to the hiring company.

 

But identifying someone at Palantir does not make that person available.

 

They may already be highly compensated. They may have valuable equity, strong colleagues, interesting work and a company brand they value. An early-stage business may be able to attract that person, but it needs to offer a compelling enough reason for the person to leave.

 

The existence of the profile is not evidence that the candidate is recruitable at the compensation, risk level or opportunity being offered.

 

That is where experienced market judgment used to play a larger role.

 

The recruiter's role is also changing.

 

Historically, an experienced recruiter could return to a hiring manager and say that the original profile made sense in theory, but the market was showing something different. Perhaps the compensation was too low for the target group. Perhaps one requirement was eliminating people with stronger evidence of actual performance. Perhaps adjacent companies were producing candidates who were better suited to solve the problem.

 

That was not a failure to execute the search. It was part of the search.

 

Today, recruiters increasingly operate inside tightly defined mandates where success is measured by how accurately and quickly they return candidates matching the specification.

 

When several recruiters are working the same mandate, the incentive becomes even stronger. The recruiter is rewarded for producing the requested profile before someone else does, not necessarily for questioning whether the profile is the best definition of the person the company should hire.

 

That changes the behavior of the recruiter.

 

Challenging a requirement can look like an inability to satisfy it. Returning exactly what was requested looks responsive, even when the first behavior may create more value for the client.

 

The result is a system that increasingly rewards compliance with the specification rather than judgment about the specification.

 

The burden then moves back to the hiring manager.

 

This is where I think the current direction becomes counterproductive.

 

One of the original purposes of using an executive recruiter was to reduce the amount of market filtering required from the hiring manager. The recruiter was supposed to understand the market, qualify candidates and narrow the field.

 

We are now moving toward a model in which increasingly powerful sourcing systems, combined with recruiters operating primarily as candidate generators, put more highly matched profiles in front of hiring managers.

 

That is obviously better than sorting through hundreds of random applicants, but it still leaves the hiring manager responsible for deciding which profiles matter, which requirements are truly important and which tradeoffs should be made.

 

The underlying structure begins to resemble an older recruiting problem. Years ago, the hiring manager might have been faced with hundreds of applications and needed help determining which ones deserved attention. Today, the list is smaller and far more relevant, but the hiring manager can still end up doing a meaningful amount of the filtering that search was supposed to remove.

 

The data suggests that some of this burden has simply moved downstream. Companies are interviewing more people per hire than they were several years ago, and technical hiring now consumes more aggregate interviewer time for each successful hire.

 

We solved a large part of the candidate discovery problem. The bottleneck shifted to evaluation and decision-making.

 

The more important question is what we should be searching for.

 

The recruiting industry is understandably focused on what AI can identify. I think the more important question is whether we are becoming equally good at defining what actually matters.

 

A job exists because a company has a problem to solve. Something needs to be built, fixed, scaled, commercialized, changed or led.

 

The purpose of a search should be to identify people whose experience provides evidence that they can solve that problem.

 

Company pedigree may be useful evidence. Education may be useful evidence. Titles, years of experience and industry exposure may all help narrow the market.

 

But those things are indicators. They should not become substitutes for determining whether the person can actually do the job.

 

The danger is that increasingly sophisticated search technology allows us to become very precise about the profile while giving us false confidence that greater precision automatically produces a better hire.

 

It does not.

 

AI has made sourcing significantly better. That is real progress.

 

But if faster sourcing encourages companies to create narrower specifications, review more candidates and place less value on recruiter judgment, then it is fair to ask whether we have improved the overall hiring process nearly as much as we think we have.

 

Executive search was never supposed to increase the number of people a hiring manager had to evaluate.

 

Its value was supposed to come from helping the hiring manager evaluate fewer people, with greater confidence that the right ones were in the room.