AI can make recruitment faster. The leadership decision is what happens to the capacity it releases and the judgement it leaves with people.

Picture the start of a recruiter’s day once AI is doing much of the first pass. The search is built. Outreach is drafted. Yesterday’s interviews have been summarised. The work that used to take the morning is waiting on the screen before the first meeting.

So is the work the system could not finish. Is that suggested match actually relevant? What did the summary miss? Why is a strong candidate hesitating? Is the hiring manager asking for the right role? Which recommendation has enough evidence to act on?

The recruiter can move faster. Their day can also become a continuous run of exceptions, judgement calls and difficult conversations. That is a very different job from the one on which many capacity assumptions were built.

Now a hiring manager calls. Someone has left, and the manager sends over the job description used to hire them six years ago. This is where the opportunity becomes interesting. The recruiter could launch the search more quickly than ever. Or they could ask whether that document still describes the work.

For a CHRO, the question goes beyond recruiter productivity. When AI changes the shape of TA work, who decides what happens to the time it releases, and what kind of contribution is TA expected to make with it?

The saved hour has an owner

Most AI business cases start with minutes removed from a process. That is useful, but those minutes do not turn into an organisational benefit on their own. The time can support more hiring volume, lower cost, stronger candidate relationships, better assessment or earlier work on capability demand. These choices may be combined, but the same hour cannot be fully committed to every one of them.

Leave the choice unmade and workload will make it for us. A faster search brings more profiles to validate. More outreach brings more replies and exceptions. The capacity target rises because the dashboard says the process is quicker. Nobody has asked whether the remaining work is more demanding or whether the recruiter has room to do it well.

A field experiment across 6,000 knowledge workers found that active AI users spent around three fewer hours a week on email, while time in meetings did not significantly change. It was not a recruitment study and should not be used to forecast recruiter capacity. It does illustrate a useful point: time saved in one activity does not automatically change the system around it.

This is why I would want the destination of the saved time stated in the business case. If the goal is throughput, say so and test what happens to decision quality and workload. If the goal is a broader TA contribution, protect time for it and give the function a mandate to use it.

What remains can be harder to carry

An hour of administration and an hour spent resolving candidate concerns, challenging a senior stakeholder or deciding whether evidence is strong enough to progress someone both occupy sixty minutes. They do not place the same demand on the person doing them.

As routine execution falls away, the human queue may become more concentrated around judgement, verification, influence and accountability. Call it decision density if the label is useful. The point is to look beyond how many tasks remain and ask how much sustained attention those tasks require.

That matters to the business as much as to the individual. If recruiters have no room between high-consequence decisions, speed may improve while rework, weak challenges and candidate frustration build elsewhere. It is possible to have a faster process and a less reliable one.

The ILO’s research on generative AI and jobs suggests that transformation of work is more likely than straightforward replacement across many exposed occupations. It does not prescribe a TA operating model. That part is a leadership choice.

Human oversight has to fit into a real day

The capacity question becomes sharper when an AI recommendation affects someone’s opportunity to work. It is easy to draw a person into a process map as the final reviewer. It is harder to show that they can understand the evidence, challenge the output and override it when necessary.

If automated recommendations grow faster than the time and capability available to review them, oversight debt builds. The approval step remains, but the quality of review weakens. Adding a sign-off box does not solve a shortage of attention or authority.

The ICO’s findings on automated recruitment give this practical weight. It found that many employers engaging in automated recruitment were likely relying on solely automated decisions. It also said that where employers use meaningful human involvement, they need to apply it consistently to candidates within a hiring stage.

A CHRO should be able to ask where human challenge is genuinely needed, how many cases will reach that point, what evidence reviewers can see, and whether they have the time and authority to disagree. Routine, reversible actions and consequential candidate decisions should not receive the same level of attention. The boundaries need to be designed, staffed and tested.

The vacancy is a design brief

Return to that six-year-old job description. The person who left may have spent a large part of their week doing work that is now automated. They may also have taken on responsibilities that never made it into the document: resolving unusual cases, connecting teams, making judgement calls when the process did not fit. A like-for-like search could recreate a role that no longer exists and miss the capability the team has just lost.

This is where the recruiter can become a Talent Architect. Start with the outcomes the team needs to deliver now. Break down the work behind them: which tasks have disappeared, which have grown, which can be shared or supported by technology, and which demand a person with particular skills and judgement. Then decide what role, if any, needs to be hired. The answer could be an external hire. It could also be an internal move, development for someone already in the team, a different mix of roles or a redesigned job.

That is not a new title to put on an old requisition process. It changes the point at which TA enters the conversation. If the role, grade and list of requirements are already approved, much of the design has been done before the recruiter arrives. The CHRO has to give TA a mandate to challenge the brief, access to people who know the work, and a way to bring workforce, skills and mobility data into the decision.

The measures need to follow the same logic. Time to fill still matters, but so does whether the new arrangement delivers the intended outcomes, whether critical work found an owner, and whether the team had to reopen the same capability gap six months later. That is a more exacting test than counting how quickly a vacancy closed.

I would want the AI business case to answer four questions in plain language. Where will the saved time go? What new human work will the technology create? Who can challenge its outputs? And will TA have the time and authority to question the work behind a vacancy before it starts a search?

AI can make recruitment faster. The more consequential CHRO decision is whether TA uses that speed to fill yesterday’s job more efficiently or to help design the work that needs doing tomorrow. If we do not decide where the time goes, the queue will decide for us.