Someone should be responsible for the workforce you actually have

In July 2024 an HR software company announced that it would put AI agents on the org chart as employees, with onboarding and performance reviews. Three days later, after hundreds of objections from HR professionals, it withdrew the plan. In 2025 Moderna merged its HR and IT departments under one executive, on the grounds that a company running more than 3,000 custom GPTs could no longer plan its people and its systems separately. Later that year Workday proposed a "Chief Work Officer," which it defined as a mindset rather than a job.

Three serious organizations, three different answers, one question none of them has settled: when the work is done by people and by autonomous agents together, who is responsible for the whole workforce?

I think the answer is an executive role with a precise charter, and I think the current answers each miss it in a revealing way.

Treating agents as employees fails because people can tell the difference and resent the pretence. An agent can be copied for the price of compute; it has no career and cannot be held responsible. Putting it on the chart dilutes the one category, personhood, on which every protection a worker has depends.

Merging HR and IT fails more slowly. It solves a real coordination problem, but it puts the people who build the systems and the people who speak for the workforce under one budget and one boss, and such a function will resolve every tension in favor of the system, because the system's costs are visible and the human costs show up later, somewhere else.

A shared mindset fails because accountability cannot be shared. When an agent gives a thousand customers the wrong answer overnight, a mindset does not get the call at six in the morning.

The role I propose is a Chief Workforce Officer: accountable for the organization's productive capacity, human and artificial, as one managed whole. The CWO decides, task family by task family, what share of the work is done by people, by agents, and by the two together, and how fast that changes. The CWO owns the rules for allocating work between them, the measurement of the whole capacity's performance, and the governance of the agents: how they are selected, deployed, evaluated, updated and retired, without ever pretending they are people. The CWO does not own the technology; that stays with the CIO, who becomes the CWO's most demanding counterpart. And the CWO reports to the chief executive, beside the CIO and the CFO, because the three of them must be able to say no to each other.

One rule inside that charter matters more than the rest. For every agent, a named human role signs for its output or the agent is not deployed. A role, held by a person with the expertise to check the work and the authority to stop it. The scarce resource in automated work is no longer the capacity to produce; it is the capacity to verify. An agent that saves ten hours and consumes eight hours of checking has earned two.

The early evidence supports the shape of this, even where it is thin. The randomized studies of generative AI at work show real productivity gains, concentrated among the least experienced workers. The first labor-market studies show almost no aggregate effect in Denmark and a sharp decline in early-career employment in the most exposed US occupations. Read together, they say that the first place automation lands is the entry layer, the routine work junior people do while learning to do the work that cannot be automated. An organization that automates that layer for five years feels only the savings. In year six it discovers it has agents nobody is qualified to check. The CWO is the only executive whose job it is to see that coming.

In most companies the Chief Workforce Officer will be today's chief human resources officer with a larger charter, and the title can wait. The charter cannot. Right now the decision about which work agents do is being made by whoever controls a software budget, and discovered by everyone else afterwards. That decision deserves an owner.