The role redesign conversation usually starts too late.
An org deploys an agent. The agent starts absorbing operational work that a human was doing. The human is still in the seat, still at the meeting, still getting paid, but carrying a narrower slice of the original job. Nobody has formally acknowledged the change or redesigned the adjacent human role around it. Three months later the human is underutilized, the agent is operating without enough human oversight, and the CHRO is trying to figure out what happened.
What happened is that the work redesigned itself and nobody was in the room to name it.
SHRM's 2026 State of AI in HR research found that AI is 5.7 times more likely to shift job responsibilities than to displace workers outright, and 3 times more likely to create new roles than eliminate existing ones. That pattern is exactly right in my experience, and it points to the thing the CHRO has to get ahead of: the shift in job responsibilities is happening whether or not anyone is managing it. The CHRO's job is to make the shift deliberate instead of accidental.
What "absorb" actually means
When an agent absorbs work, it does not absorb a whole job. It absorbs a layer of the job. Specifically, it absorbs the operational execution layer: the scanning, pulling, flagging, drafting, routing, and logging that previously required a human to sit down and do it, step by step, on a schedule.
What it does not absorb is judgment, relationship, escalation, and accountability. Those stay human not because agents cannot simulate them, but because an agent cannot be accountable for the downstream consequences of judgment calls. MIT SMR's research makes this explicit: "Agentic AI cannot be accountable for its decisions. The deploying human is."
The redesign question is not "will this agent take this person's job." The question is "when this agent takes the operational layer, what does this person's job become?" That is a specific, answerable question for each seat on your chart. And it requires someone in the CHRO role to sit down and answer it before the agent is deployed, not after.
The before and after at Sneeze It
I will make this concrete. Here is what role redesign actually looked like across several seats when we built out our hybrid team.
Before Radar: The first forty minutes of my day were operational. Pulling Slack threads, checking the calendar, scanning the inbox, noting open delegations, assembling a picture of what needed attention. I was the synthesizer and the decision-maker in the same pass. The cognitive work of scanning was consuming the time I needed for the decisions that required actual judgment.
After Radar: Radar runs the morning scan. By the time I sit down, there is a compiled briefing covering Slack, calendar, pipeline, inbox, project status, and open delegations. My first forty minutes became something different: reviewing a summary and making decisions. The scan moved to the agent seat. The judgment stayed human. My role did not shrink. The operational part of it moved to Radar and what remained was higher-value work.
Before Dash: Our account managers were spending meaningful time each week pulling data from Meta and Google ad accounts, looking for spend anomalies, checking CPL trends, flagging issues to clients. It was manual, it was repetitive, and it happened inconsistently because people were busy.
After Dash: Dash reads thirty-plus client accounts daily and surfaces anomalies automatically. The account managers receive the pattern report and decide what to do with it. The data pull moved to the agent seat. The client relationship, the interpretation, and the decision about what to tell the client stayed human. Deloitte's 2025 Global Human Capital Trends found that managers spend roughly 40% of their time on administrative work versus 13% on people development. Dash did not eliminate the account manager role. It shifted the ratio.
Before Arin: We had a human call center manager doing the same analysis Arin now handles at the data layer: reading appointment rates by caller, tracking speed-to-lead, identifying who was below target, drafting feedback. It was happening once a day at best, and the quality of the feedback was inconsistent because the manager was also doing everything else managers do.
After Arin: Arin reads the CCM Stats template daily and produces coaching drafts for our callers Amanda and Erica. I review and approve before anything goes out. The first-pass analysis moved to the agent seat. The coaching judgment, the relationship with the caller, the decision about tone and timing: those stayed human. Arin also created something new: the human coaching role became more focused on the calls that actually required human judgment, because Arin was handling the ones that were purely data-driven.
Before Tally: Pushing KPI values to our scorecard was a recurring manual step that fell to whoever remembered. It happened inconsistently. The scorecard was frequently stale.
After Tally: Tally pushes KPI values four times a day. The scorecard is current. Nobody carries that task anymore. What the humans who previously tracked KPIs manually now carry is what to do with the current numbers, not the work of producing them.
In each of these cases, the human role did not disappear. It shifted upward. The operational execution layer moved to the agent. The judgment layer, the accountability layer, and the relationship layer stayed human.
The accountability split that makes this work
Here is where the research tension matters and where I want to be direct about it.
The 2025-26 literature is split on how to think about agents. Camp A, represented by MIT SMR and HBR's "agent manager" framing, says agentic AI should be managed more like a coworker than a traditional tool, with dashboards, scorecards, and observability. Camp B, represented by HBR and BCG research published in May 2026, says that anthropomorphizing agents in practice produces measurable harm: reduced individual accountability among humans, increased unnecessary escalation, lower review quality. Their recommendation is to treat agents as "rented contractors with a narrow statement of work," governed by scoped permissions, kill switches, and named human owners.
Both camps are right about the substance. Every agent needs a named human owner. Every agent seat needs a measured output tied to a business outcome. Human accountability must never migrate to the agent.
Where they differ is on framing, and the framing matters for role redesign.
When Arin's coaching draft goes to Amanda, Amanda knows a human reviewed it. When Dash flags an anomaly, the account manager knows a human will decide what to do with it. The agent holds the operational layer. The human holds the judgment layer. That split is explicit, designed, and maintained on purpose. The accountability never moved to the agent because the accountability architecture was built before the agent was deployed.
In April 2026, we retired Jeff, one of our agents. Jeff had held a data integrity seat that was originally scoped clearly. Over time, the work Jeff was built for migrated into other seats: Dash absorbed the ad monitoring work, Dirk absorbed the revenue signal work. Jeff's seat stopped being needed. A human decision, made through a formal review process, with capabilities explicitly reassigned to named human-owned seats. Jeff was not managed like an employee. Jeff was managed like a seat that either earns its place on the chart or does not. When it stopped earning its place, a human retired it. That is the accountability architecture working correctly.
"Onboarding" an agent means defining scoped permissions and a clear metric before deployment. "Retiring" an agent means a human decides the seat is not earning its place and redistributes the work. Neither requires HR onboarding procedures designed for human employees. Both require the same structural discipline: named owner, measured seat, clear accountability chain.
What the CHRO owns in this redesign
Role redesign around agent absorption is not a spontaneous outcome. It is a deliberate act that requires someone with the authority and the cross-functional view to do it. That is the CHRO.
Korn Ferry found that 42% of CHROs are already prioritizing AI investment but only 5% feel fully prepared. The gap is partly a framing problem. The CHROs who are uncertain are often waiting for a clear set of procedures for "managing AI agents." The procedures they need are not AI-specific. They are the accountability architecture procedures that any good workforce designer already knows: define the seat, name the owner, measure the output, retire what does not work.
What is new is the scope. When agents absorb the operational layer across multiple seats simultaneously, the human roles adjacent to all of those seats change at the same time. The CHRO needs to map that change explicitly: here is what the analyst seat looked like before, here is what it looks like now that Dash holds the data pull, here is what the skill profile for a human analyst in this structure actually requires.
That mapping does not happen automatically. The agent deploys, the work shifts, and if nobody has mapped the human role around the shift, the human is underutilized and the accountability structure is unclear. HBR Analytic Services found that only 6% of leaders fully trust agents with core processes. The reason trust is low is usually not the agent's capability. It is the absence of a clear accountability architecture around what the agent owns and what the human retains.
The mission is to let agents carry the operational work, so people are free for the work that matters. The CHRO delivers that mission by mapping the before and after for each seat adjacent to an agent deployment, naming the boundary explicitly, and building the accountability structure that keeps humans in charge of the judgment layer. That is the work. It is not anthropomorphizing agents. It is redesigning roles so the humans in them are doing the work humans are actually needed for.
That redesign has to happen before the agent goes live, not after.
See the live chart
Every seat on the Sneeze It org chart is queryable through the OTP MCP, showing which seats are agent-owned versus human-owned and what outputs each seat is accountable for.
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 tell me which seats are held by agents and what the named human owner is for each agent seat."
You will see the accountability structure of a running hybrid team, not a diagram of how one could theoretically work.