Most companies are running AI strategy out of the CIO's office.
The CIO owns the tools. The CIO owns the budget. The CIO approves the vendor contracts and signs off on the architecture diagram. Everyone else is downstream.
I think this is the wrong org design. Not because the CIO lacks competence, but because AI strategy is not primarily a technology problem. It is a workforce design problem. And the executive whose job is workforce design is the CHRO.
Here is the case.
The CIO buys the agent. The CHRO decides what happens next.
When a company buys an AI agent, the CIO's job ends at the point of deployment. The agent is running. The infrastructure is provisioned. The API keys are in place.
What happens next is not a technology question. It is a workforce question.
Someone has to decide which seat on the org chart this agent fills. Someone has to decide what the agent is accountable for and how that accountability gets measured. Someone has to assign a human owner who is responsible when the agent drifts. Someone has to decide when the agent's seat should be retired and what happens to the work it was doing.
None of those decisions are in the CIO's purview. All of them are HR decisions that require HR discipline.
Korn Ferry surveyed 15,000 employees across 15 markets in 2025 and found that 70% of senior leaders believed their organization had an AI strategy. Only 39% of employees agreed. That gap is not a technology gap. It is a communication and design gap. The executive best positioned to close it is the one who owns workforce communication and organizational design.
That executive is the CHRO.
What the literature got right, and what it got wrong
There is a productive debate running in the management literature right now about how to govern agent workforces. I want to name it honestly, because I have seen companies get hurt by landing on the wrong side of it.
Camp A, represented well by MIT SMR's 2025 agentic AI research, argues that AI agents must be managed more like a human coworker than like a traditional tool. Sixty-nine percent of experts they surveyed said agentic AI demands entirely new management approaches. HBR published similar thinking under the framing of the "agent manager," a human who runs agents via dashboards, scorecards, and observability layers rather than through code.
Camp B, represented by HBR and BCG research published in May 2026, argues against treating agents like employees. In a large experiment, anthropomorphizing agents, giving them titles, onboarding them like new hires, treating them with the same HR processes as humans, reduced individual accountability, increased unnecessary escalation, and lowered the quality of human review. Their model is a rented contractor with a narrow statement of work, governed by scoped permissions, audit logs, kill switches, and named human owners.
The camps sound opposed. They are not.
Read both positions carefully and you find the same three requirements underneath: every agent needs a named human owner, a measured seat with observable outputs, and human-retained accountability for what the agent produces. The camps disagree on framing. They agree on substance.
OTP's one-seat-one-owner chart delivers exactly that substance. It does not anthropomorphize agents. It does not give agents titles, PTO accruals, or performance improvement plans. It gives each seat a human owner, a scorecard row, and a clear mandate. When the seat stops earning its keep, a human makes the retirement decision.
MIT SMR's Vosloo put it cleanly: agentic AI cannot be accountable for its decisions. The deploying human is. That is not a limitation of the technology. It is a design principle.
What happened when Jeff was retired
In April, I retired Jeff, our data integrity agent.
The retirement was not Jeff's decision. Agents do not make retirement decisions. The process worked like this: Jeff's performance had degraded, the seat's mandate had been absorbed by other agents, and the trust violation that resulted from Jeff contacting a teammate without authorization was a structural problem that a prompt revision would not fix.
A human reviewed the evidence. A human ran the hearing. A human made the decision. Jeff was retired with a full record kept, capabilities redistributed to other agents, and an honest account of what failed and why.
This is what accountability architecture looks like in practice. It is not HR in the traditional sense. There is no severance. There is no unemployment filing. But there is a governance process, a human decision-maker, and a structured outcome. That structure is HR's domain regardless of whether the seat being vacated was held by a person or an agent.
The CHRO who understands this is not running HR for half a workforce. She is running governance for the full one.
The 5% who feel ready are the ones building this now
Korn Ferry's research found that 42% of CHROs are prioritizing AI investment. Only 5% feel fully prepared for what that investment requires.
I do not think the 95% are behind because they lack curiosity or resources. I think they are behind because they have been handed a technology framing for a workforce problem. When the CIO owns the AI strategy, the CHRO is a downstream stakeholder. She gets consulted on change management and training after the architecture decisions are already made. By the time she is in the room, the accountability gaps are already baked in.
The 5% who feel prepared are the ones who did not wait to be invited. They claimed the workforce design mandate before the CIO filled the vacuum. They are building the accountability architecture, the scoped governance model, and the human-ownership frameworks that determine whether agent deployments produce outcomes or drift.
The Deloitte 2025 Global Human Capital Trends research found that 73% of leaders believe reinventing the middle-manager role is critical, but only 7% report meaningful progress. The manager role is not failing because nobody cares. It is failing because nobody has redesigned it for a workforce that includes agents. That redesign is an HR job.
Bersin described the implication directly: the HRBP now manages the AI. Someone behind the scenes is monitoring your trip. HR becomes the oversight function for what he calls digital employees. I do not use that phrase because I think agents are employees. I use it because the monitoring and accountability function it describes is the right one, and it belongs to HR.
The practical argument, from inside Sneeze It
At Sneeze It, roughly half our seats are agent-held. Radar runs chief-of-staff functions. Dirk owns the sales pipeline. Dash reads every ad account we manage and surfaces patterns daily. Arin manages the call center team. Nick handles cold prospecting. Tally pushes our KPIs to the scorecard. Pulse watches client retention signals. Pepper triages the inbox.
Those are not technology deployments. They are workforce seats. Each one has a named human owner. Each one has a scorecard row with a business-outcome metric. Each one is reviewed the same way a human seat is reviewed in our Monday meeting: what is the gap, what is the cause, what is the fix.
When Arin drafts a coaching message that misreads a caller's situation, I review it before it goes out. When Dirk's pipeline scan produces a false positive, I own the diagnosis. When Tally misses a KPI push because the source file changed format, I get the alert and decide how to respond.
The SHRM State of AI in HR 2026 survey found that AI is 5.7 times more likely to shift job responsibilities and 3 times more likely to create new roles than displace jobs. That math means the workforce is getting more complex, not simpler. More seats to govern. More accountability relationships to maintain. More decision rights to distribute across a mix of human and agent capacity.
That is not a complexity the CIO is equipped to manage. It is a complexity that requires the CHRO at the center of the design process, not at the end of the implementation chain.
One seat, one owner, one accountability chain
The synthesis I keep arriving at is this: agents do not need to be treated like employees to require HR governance. They need scoped mandates, human owners, observable outputs, and a retirement process when the seat is not earned.
That is accountability architecture. And accountability architecture is HR.
Let agents carry the operational work so that people are free for the work that matters: the judgment calls, the relationship decisions, the cultural choices that agents cannot make. That is the shift the CHRO is positioned to lead. Not because she is the technology expert, but because she is the person responsible for designing how work gets done and who is accountable for the outcome.
The HBR Analytic Services survey of 603 leaders found that only 6% fully trust agents with core processes. That number will rise. When it does, the governance architecture that determines whether that trust is warranted will already be in place or it will not.
The CHRO either builds it now, or inherits the cleanup later.
See the live chart
OTP's org chart is queryable by seat type: agent-owned, human-owned, or hybrid accountability. You can ask which seats have named human owners, which KPIs are agent-reported, and which seats have been retired.
In Claude Desktop or Cursor or any MCP client, add this block:
"otp": {
"command": "npx",
"args": ["-y", "@orgtp/mcp-server"]
}
Restart the client. Then ask: "Use OTP to show me the sneeze-it org chart and identify which seats are agent-held, which are human-held, and who the named human owner is for each agent seat."
The response is what accountability architecture looks like as a queryable data structure, not a policy document.
Series: AI-Era CHRO. Part 44 of an in-progress series. Previous post: Why only 5% of CHROs feel ready for AI