The first time I put an agent on our org chart, I did not think of it as a recruiting decision.
I thought of it as a deployment. I had a workflow that needed coverage. I spun up an agent. I connected it to the right tools. Done.
Six weeks later the agent was running confidently and the outcome I cared about had not moved. I went looking for the gap and found it in the same place every bad hire goes wrong: nobody had defined what the seat was actually for, what it needed to produce, and who was accountable when it did not.
That is a recruiting failure. It just happened at the speed of a deployment instead of the speed of a job search.
Here is what I know now: when agents can hold seats on your org chart, recruiting does not get smaller. It gets a second track. You are filling two kinds of seats, and the discipline for each is different. Getting that discipline wrong on the agent track costs you differently than getting it wrong on the human track. But the cost is real either way.
What recruiting is actually for
Recruiting is how organizations match capacity to work. You identify a gap. You define the seat. You find the entity that can fill it. You integrate that entity into the operating system. You hold the seat accountable.
Every word in that description applies equally to a human hire and to an agent deployment. The mechanics differ. The discipline is the same.
SHRM's 2026 State of AI in HR report found that AI is 5.7 times more likely to shift job responsibilities and 3 times more likely to create new roles than to displace jobs outright. The implication for recruiting is underappreciated. If your org is gaining new roles faster than it is losing old ones, your recruiting function is not shrinking in response to agents. It is expanding to cover seats of a new type.
The teams that treat agent deployment as a purely technical function, separate from how they think about hiring, are the teams that end up with agents nobody owns. You can see this pattern in the trust data. HBR Analytic Services surveyed 603 business leaders in December 2025 and found that only 6 percent fully trust agents with core processes. I do not think that low number reflects a technology problem. I think it reflects an accountability gap. Agents got deployed without the recruiting discipline that would have made them trustworthy.
The two seats are sourced differently
When I am filling a human seat, I am looking for a person with a specific skill profile and judgment pattern. The sourcing process has established infrastructure: job descriptions, interviews, references, offers, onboarding.
When I am filling an agent seat, I am doing something structurally different. I am defining a scope of work, identifying the toolset the agent will need, specifying the observable output that will tell me whether the seat is working, and naming the human who will own the seat's accountability. Those four things are the agent version of a job description and a reporting structure. Without them, the deployment is not a hire. It is an experiment with no outcome criteria.
At Sneeze It, every agent on our chart was sourced against this checklist. Radar covers chief-of-staff functions: calendar, briefings, Slack monitoring, cross-channel synthesis. The output that tells me the seat is working is a daily briefing that surfaces what I need to know and flags what I need to decide. Tally pushes KPI values from local data sources to the scorecard on a four-times-daily schedule. The output is simple: the numbers are current or they are not. Dash reads every ad account we manage across Meta and Google and reports patterns daily. The output that tells me the seat is working is an alert within four hours when a client account moves outside baseline.
Each of those seat definitions came before the agent was deployed. The seat definition is the recruiting artifact. Without it, you are not doing hybrid recruiting. You are doing reactive triage.
The accountability question is the recruiting question
Here is where the literature gets into a legitimate argument, and I want to handle it directly.
One school of thought, represented in MIT SMR's work and in HBR's writing on "agent managers," says agents should be managed more like coworkers than like traditional software. Sixty-nine percent of experts in MIT SMR's 2025 agentic AI research agreed that new management approaches are required. The argument is that agent seats need dashboards, scorecards, and observability, and that humans need to develop new skills around supervising autonomous AI.
A competing school of thought, published in HBR by a team that includes BCG researchers in May 2026, says the opposite: do not treat agents like employees. Their empirical finding was that anthropomorphizing agents in organizational settings reduced individual human accountability, increased unnecessary escalation, and lowered the quality of human review. Their recommendation was to treat agents as a "rented contractor with a narrow statement of work," governed by scoped permissions, kill switches, audit logs, and named human owners.
I have run agents long enough to know both camps are right about the substance and arguing about the framing.
The MIT SMR camp is right that you need dashboards, scorecards, and new human skills around supervision. The HBR/BCG camp is right that the agent itself cannot be made accountable. MIT SMR's own Vosloo makes this explicit: "agentic AI cannot be accountable for its decisions." The deploying human is.
The synthesis is not complicated. The agent gets a seat with a named human owner and a measured output. The human owner is accountable. The recruiting question is not "should we treat this agent like an employee?" The recruiting question is "who is the human that owns the accountability for this seat, and does that human understand what they are taking on?"
At Sneeze It, the answer to that question is always me or Bogdan, our COO. When Arin, our call center manager agent, drafts a coaching message for the calling team, I review it before it goes out. The agent produces. The human decides and owns. That is not anthropomorphizing. That is accountability architecture.
When I retired Jeff, our former data intelligence agent, in April, the retirement was a human decision made through a structured hearing. Jeff held a seat. The seat was not earning its keep. The capabilities were redistributed to Dash and to Dirk. Jeff's retirement was documented. The accountability for what had been Jeff's seat transferred explicitly. That is what clean agent offboarding looks like, and it only happens when recruiting was clean going in.
What changes about sourcing humans
Here is what agents have actually done to how I think about hiring humans.
When an agent absorbs the operational toil of a seat, the human version of that seat needs to move up the judgment stack. This is the Bersin framing, which I find accurate: for every dollar spent on machine learning technology, companies spend roughly nine dollars on human capital to integrate and govern it. The human capital does not go away. It reorganizes toward judgment, oversight, and the decisions that cannot be automated.
Deloitte's 2025 Global Human Capital Trends research found that managers currently spend about 40 percent of their time on administrative work and only 13 percent on actual people development. Agents can absorb much of that 40 percent. But the skills a manager needs to develop people, resolve conflict, and hold culture are not skills that get built by default. They need to be recruited for explicitly.
When I think about hiring a human seat at Sneeze It now, I am thinking about the agent layer that will sit below that seat and asking: what does this person need to be excellent at, given that the coordination, scheduling, data preparation, and first-draft work will be handled by agents? The job description changes. The sourcing criteria change. The interview questions change.
This is not a smaller recruiting function. It is a more demanding one.
The governance gap is a recruiting gap
Korn Ferry surveyed 15,000 employees across 15 markets in 2025 and found that 70 percent of senior leaders say their organization has an AI strategy, but only 39 percent of employees agree. That 31-point gap is not primarily a communication problem. It is a recruiting and design problem. The seats that would close that gap, human and agent, have not been filled. The accountability owners for the agent layer have not been named. The scorecards that would make the strategy visible to employees have not been built.
Forty-two percent of CHROs say they are prioritizing AI investment for HR, but only 5 percent feel fully prepared, per Korn Ferry's 2025 CHRO research. The preparation gap is partly a knowledge gap. But it is mostly a seat-definition gap. CHROs who feel unprepared are, in almost every case, CHROs who have not yet defined what the agent seats on their chart are supposed to produce, who owns them, and how they will be measured.
That is a recruiting problem with a recruiting solution.
Let agents carry the operational work
The underlying reason to get hybrid recruiting right is not efficiency. It is role clarity.
Every human seat that carries operational toil it should not be carrying is a seat where the person is not doing the work they were actually hired for. When Radar handles the daily briefing, when Tally keeps the scorecard current, when Dash surfaces the anomalies, the humans on our chart are not doing that work. They are doing the judgment work that compounds: strategic decisions, client relationships, coaching, the calls that require a person.
SHRM's finding that AI creates roles three times as often as it displaces them points toward an org chart that is growing in scope, not shrinking. But the growth only compounds if the new seats, human and agent, are filled with the same discipline that good recruiting has always required.
The seat definition. The output criteria. The named human owner. The scorecard that makes accountability visible.
Those four things have not changed. The number of seats that need them has doubled.
See the live chart
Every agent and human seat on our chart is queryable from OTP's MCP server, including which seats are agent-owned versus human-owned and what each seat's measured output is.
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 which seats on the sneeze-it chart are held by agents versus humans, and what each agent seat is measured on."
The response shows you what hybrid recruiting accountability looks like when it is explicit rather than assumed.
Series: The AI-Era CHRO. Post 35. Previous in series: HR does not disappear when half your workforce is agents. It changes shape entirely.