Share with your CHRO
Every CHRO Josh Bersin talks to is being asked to automate up to a third of their HR headcount, and Bersin’s firm is building toward roughly 10 “super agents” that can run common HR automation workflows end-to-end without human handoffs. Gartner’s Chris Hester offers a cooler read: 64% of HR leaders report AI productivity gains, but only 25% say costs have actually dropped, suggesting most AI investment is hitting low-value tasks. The interoperability question, specifically whether MCP and Google’s A2A protocol can connect agents across platforms, remains genuinely unsettled.
What this means for your business
The gap between “productivity gains” and “cost reduction” is the most diagnostic number in this story. If your HR AI investments are producing the former without the latter, you’re automating the wrong things. CHROs sitting on pilot programs and no formal strategy, which describes about 80% of Gartner’s client base, are accumulating technical debt in the form of disconnected point solutions that will be harder to stitch together once the architectural standards settle.
Bersin is an analyst who sells advisory services and whose firm is actively building the super-agent products he’s describing, which tilts his timeline toward urgency and his confidence toward the optimistic end. That doesn’t make him wrong. His core argument, that citizen AI experimentation needs to give way to top-down architectural planning, is sound and matches the pattern visible in every prior enterprise technology wave from ERP rollouts to cloud migration. The organizations that waited for the “right time” to consolidate their data infrastructure in 2015 spent 2020 scrambling. The dynamic is familiar.
The interoperability problem is the actual gating factor here, not AI capability. Agents that can’t pass context reliably across HR, finance, and IT systems can’t run full workflows autonomously, full stop. If AWS, Google, and Microsoft converge on a working A2A standard by late 2025, the CHRO’s 2026 budget conversation shifts from “should we invest” to “which vendor’s agent architecture do we standardize on.” That’s the decision worth pressure-testing now, before a platform bet gets locked in during a renewal cycle.
Concept deep-dive: Agentic AI
Agentic AI refers to software that doesn’t just answer questions but takes sequences of actions autonomously to complete a goal, think of it as the difference between a calculator and an intern who knows when to escalate. In HR, a single agent might screen candidates, schedule interviews, and update the ATS without a human touching each step. The business relevance is that agent reliability depends entirely on clean integrations between systems, which is why interoperability standards are the actual constraint, not the AI models themselves.
Based on reporting from This could be the year AI automation takes over HR, originally published 2026-01-21 03:00:00.
