The Josh Bersin Company, a well known research firm that advises corporate HR departments, released a report in June projecting that HR headcount could shrink by 30 to 50 percent by 2030. The driver is AI agents, which are software programs that can complete a specific task on their own, like answering a worker's question about their pay stub or updating payroll when tax rules change.
The report describes a future built on more than 100 different specialized AI agents running across HR, each handling a narrow slice of the work: recruiting, benefits, compliance, learning, and more. Bersin's own team is careful to call this a restructuring rather than a wipeout. They even warn against "agent sprawl," meaning companies should not just bolt on dozens of disconnected AI tools and call it a strategy. The goal is a redesigned HR setup, not a smaller version of the old one running on autopilot.
That framing matters, because the headline number, half of HR jobs gone, is easy to misread as an announcement that companies are actively cutting staff right now. They are not, according to Eser Rizaoglu, an analyst at Gartner who covers HR. He points out that the layoffs happening in HR departments today are mostly a correction for over-hiring during the pandemic, not a direct result of AI replacing people.
Bersin backs this up with a simple data point: job postings for HR roles have stayed flat rather than dropping. That looks less like a profession disappearing and more like a profession getting rearranged.
The practical reason full automation is not happening yet is less exciting than the AI headlines suggest. Someone has to teach these agents how a specific company actually works, including where the data lives, which processes are genuinely repetitive, and which ones only look repetitive on paper. Bersin describes this as an ongoing training relationship, not a one time software purchase. Older HR software could be bought and used right out of the box. AI agents need continuous coaching to function inside the specific, often messy way a real company operates.
There is also a quieter risk that gets less attention: institutional knowledge. When an AI agent makes a mistake, and it will, someone with real knowledge of the company's workflow has to catch it and fix it. If a company cuts the staff who hold that knowledge, nobody is left who can tell whether the AI got something wrong in the first place.
Gartner's survey of HR leaders from earlier this year found that the biggest obstacles to using AI agents were not really about the AI itself. They were about messy internal data, systems that do not talk to each other, insufficient staff training, and plain resistance from employees who do not trust the tools yet. That is a much less dramatic story than mass layoffs, but it is probably the more accurate one for now.
For any business leader watching this trend, the lesson is not that HR jobs are safe or doomed. It is that the roles most exposed are the ones built entirely around repetitive, rule following tasks. The roles most protected are the ones that involve judgment calls, handling sensitive situations, or knowing how the pieces of a company actually fit together. That pattern will likely repeat in every other department eyeing AI agents next, not just HR.