Share with your CHRO
McKinsey’s people and organization practice is making a direct case that agentic AI, meaning AI systems that plan and act autonomously rather than just answering questions, forces HR to stop being a support function and become the architect of how humans and software agents divide work. The argument covers six domains: workforce planning, org design, hiring, learning, talent management, and employee experience. The throughline is that static role-based HR operating models break when the unit of work shifts from “job” to “task,” and HR’s redesign mandate is the first enterprise response that actually has to land.
What this means for your business
The CHRO who reads this as a flattering endorsement of HR’s rising status is reading it wrong. The actual claim is threatening: if HR doesn’t redesign its own operating model around task-level human-agent allocation, some other function, likely the CIO or a newly minted Chief AI Officer, will do it instead, and HR becomes a policy attachment to someone else’s architecture. The organizations where that power struggle is already underway are the ones this piece is written for.
The most practically loaded idea here is the shift from role-based to activity-based workforce planning. Today, headcount plans answer “how many people in which roles.” An activity-based model answers “which specific tasks should a human own, which should an agent own, and which need both.” That sounds like a modeling exercise, but it’s actually a classification problem that touches every job description, every performance framework, and every compensation band in the company. HR teams that haven’t started mapping their workflows at task granularity don’t have a planning gap, they have an architecture gap, and closing it requires data infrastructure HR rarely controls today.
McKinsey, whose advisory model profits from helping enterprises build exactly the hybrid workforce systems it describes here, has an obvious incentive to present the transformation as both urgent and HR-led rather than, say, IT-led or deferred. That tilt probably compresses the realistic timeline and understates how hard it is to get HR the data access required to do activity-level workforce modeling in real time. The falsification condition is simple: if the companies winning at agentic deployment in the next two years are the ones that let engineering or operations lead the task-allocation design rather than HR, this framing collapses. That outcome is at least as plausible as the one described.
Concept deep-dive: Activity-based workforce planning
Traditional workforce planning counts bodies in boxes, a headcount number per role per team. Activity-based planning goes one level deeper and asks which discrete tasks make up each role, then assigns each task to the best performer, human or AI agent. Think of it like a bill of materials in manufacturing, but for work itself. The business connection is direct: you can’t make a rational build-versus-buy-versus-automate decision until you know what the actual work units are, not just what the job titles say.
Based on reporting from HR’s transformative role in an agentic future, originally published 2025-11-17 03:00:00.
