ClickUp Cut 290. "Agent Manager" Returns UnitedHealth on LinkedIn.
ClickUp laid off 290 on May 21 and coined "Agent Manager" as the new role. LinkedIn boolean returns 127 insurance supervisors. Source the real pool now.
On May 21, 2026, ClickUp CEO Zeb Evans posted to his 229,600 X followers that he was cutting 22% of the company, about 290 of 1,300 employees, and in the same memo coined a new title: "agent manager." Days earlier, Fortune had revealed ClickUp runs roughly 3,000 internal AI agents, a 3:1 agent-to-human ratio, with one growth ops manager publicly overseeing 37 of them. The 290 people who just left are the most agent-experienced free agents in the U.S. labor market right now, and your LinkedIn boolean cannot find them.
The title Zeb Evans just invented does not exist in any database yet
Read the memo carefully. Evans did not write "AI operations lead" or "automation specialist." He wrote that employees who automate their own jobs "become owners of the AI systems, agent managers." That phrase is the first CEO-level coinage of "agent manager" as a knowledge-work role, and it landed with the force of a $4B SaaS founder telling his industry what to call the job.
Now run the search. Across U.S. professional-network data, the exact title "Agent Manager" returns about 127 profiles. The top employer is UnitedHealth Group with 16. Then AT&T. Then Intracoastal Realty. Then HealthMarkets. Then Baptist Health Plan. Every one of them is an insurance agent supervisor or a real estate franchise lead. Not one is orchestrating AI.
This is the boolean trap. Recruiters running "Agent Manager" AND AI over the next 90 days will pull a list of zero ClickUp leavers and a long tail of insurance ops. Even worse: when the 290 do start updating their headlines, they will be buried under the 100,000-plus existing insurance and realty agent managers already squatting the phrase.
Title-based search is structurally broken for AI-native roles
The Evans memo is one data point in a larger pattern we keep documenting on this blog. GitHub shipped .agent.md and LinkedIn returned 14 people. "AI Agent Engineer" as a self-described title? 93 profiles, while awesome-copilot has 346 contributors. "Vibe coding" shows up in Adobe and JPMorgan JDs and boolean returns 7. The vocabulary moves faster than the index, and the index is what your ATS and your LinkedIn Recruiter seat both run on.
The fix is not a better boolean. It is sourcing by behavior, not by title. Which is exactly why we built Refolk: you describe the person in plain English ("growth ops or PM who managed 20-plus AI agents at a SaaS company in the last 12 months") and you get a ranked shortlist, including profiles whose headlines still say "Senior Growth Manager" because they were laid off last Thursday and have not updated yet.
Why the 290 leavers, not the survivors, are the trade
Most layoff coverage gets this backwards. The press wrote up the $1M cash salary bands Evans announced for the people who stayed, and recruiters started DM-ing survivors. That is the wrong move. The survivors are price-anchored to seven-figure cash and will not move for less. The leavers are the actual hireable pool, and they are uniquely valuable for three specific reasons.
First, they have been in ClickUp's agent-native operating model since January 2026. Four months of daily practice managing fleets of agents across Slack, code review, customer support, and product workflows. A KPMG Q1 2026 report found only 9% of businesses orchestrate multiple AI agents across workflows. The ClickUp 290 are in the top decile of practitioners by experience, by a wide margin.
Second, ClickUp acquired Codegen in December 2025. Anyone touching engineering or product at ClickUp in the last five months has hands-on time with production-scale agentic codegen, a tool most competitors are still piloting on a Friday afternoon.
Third, Evans's own words filtered who survived. From the memo: "AI makes the best engineers wildly more productive, and everyone else using AI slows these engineers down." That is a brutal sentence, and it tells you the cut was not random. The 290 include people who were not slow, just not in the top tier of a top-tier-only company. Most engineering orgs in the U.S. would consider them excellent hires.
The survivors are locked in by million-dollar cash. The leavers are the most agent-fluent operators on the open market.
The named human signal: Andy Cabasso and the number 37
Andy Cabasso, ClickUp growth ops manager, told Business Insider he personally oversees 37 AI agents. That is the canonical public reference for what an "agent manager" actually does. It is also the kind of signal that does not live in a job title. It lives in About sections, conference bios, personal sites, GitHub READMEs, and Loom intros.
If you are sourcing right now, the heuristic is: anyone publicly claiming responsibility for more than 10 agents is operating at ClickUp or Polsia tier. Polsia, by the way, just raised $30M at a $250M valuation as a one-person startup running software ops for solopreneurs. Ben Cera is the logical extreme of the Evans thesis, and he is also a competing buyer for the same 290 people you want to hire.
Geography: ignore the San Diego headline
ClickUp's HQ is San Diego. Naive recruiters will geo-filter there and miss almost everyone. The actual ClickUp talent footprint, roughly 1,240 current and former affiliated profiles in U.S. professional-network data, clusters in San Francisco, the broader Bay Area, Portland Oregon, New York City, and Nashville. The 290 cut on May 21 are distributed across all of those metros and more.
This is the second place natural-language sourcing matters more than boolean. A San Diego filter would have caught maybe a quarter of the pool. A Refolk query that asks for "ex-ClickUp staff in any U.S. metro who shipped agent orchestration work" returns the whole footprint, weighted by recency.
The three role archetypes Evans named, and who buys each
Evans was unusually specific about the post-headcount org. He described three archetypes. Treat them as three different hiring funnels, because the buyers are different.
Builders
Engineers and PMs directing code-writing agents. These are the people with Codegen exposure. Direct competitors here are GitLab (which publicly restructured for the "agentic era" the same week), Cursor, Anthropic, and any Series A or B shipping developer tools. They will move for technical leverage and equity, not cash. The $1M cash bands at ClickUp will not pull them back once they have left, because the bands were never on offer to the 290 in the first place.
System Managers
Workflow automators. Cabasso sits here. These are the operators a Series A founder dreams about: one person who can replace a five-person ops team because they know how to wire up agents across Slack, HubSpot, Linear, and Stripe. Competing buyers are Relay.app, Notion, Asana, Monday.com, and every YC company in the current batch. ServiceNow and Salesforce Agentforce are slower but have the budget.
Front-liners
Customer-facing roles that survived because they are the human interface on top of agent fleets. These are the hardest to poach because they are the most visible inside ClickUp and the most likely to be on monthly performance awards. ClickUp has issued about five monthly performance awards since 2022, with criteria explicitly shifted in the past year to reward AI leverage. If you are recruiting here, do not bother with cash. Pitch autonomy.
The 100x band is a piece-rate model dressed as a salary
One last thing worth saying out loud. The $1M cash band is not a SaaS engineering ladder. It is closer to law firm partner draws or trading floor PnL. Evans is paying for output, not seniority, and he has said so. That filter changes who stays and who leaves.
The people who stayed are output-obsessed and think like owners. The people who left, in many cases, were career-band climbers who did fine at a normal SaaS company but did not fit a piece-rate culture. That is a feature for most founders hiring from this pool. You are getting operators who tried to make the new model work, learned the playbook, and now want a more reasonable cadence. Pitch them ownership, scope, and a clear scoreboard, and you will close.
The 60-day window
The standard layoff sourcing window is 60 to 90 days before the pool saturates. ClickUp is different. The vocabulary problem buys you longer, maybe 120 days, because most recruiters will spend the first month running broken boolean and the second month complaining that "there's no one out there." There is. They are not calling themselves agent managers yet, and when they do, you will be racing the entire insurance industry for the phrase.
Two practical moves this week. Build a natural-language query (in Refolk or anywhere else that supports it) that targets behavior, not title: "managed AI agents," "automated my role," "shipped with Codegen," "ex-ClickUp." Then layer in the GitHub signal. ClickUp survivors and leavers with engineering exposure will have public repos referencing agent orchestration patterns, MCP servers, or codegen workflows from the last six months. That is the cleanest filter you will get.
FAQ
How many people did ClickUp actually lay off on May 21, 2026?
Roughly 290 employees, or 22% of a 1,300-person headcount. The cut was announced by CEO Zeb Evans directly on his personal X account, not through an internal all-hands, which is itself a signal about how the company now operates. The same memo introduced $1M cash salary bands for "100x impact" survivors and coined "agent manager" as the new role title for employees who automate their own jobs.
Why does searching "Agent Manager" on LinkedIn return insurance employees?
Because the title predates Evans's coinage by decades in the insurance and real estate industries, where it refers to supervisors of human sales agents. About 127 U.S. profiles use the exact title today, with UnitedHealth Group, AT&T, Intracoastal Realty, HealthMarkets, and Baptist Health Plan as the top employers. None of those people orchestrate AI. Standard recruiter boolean cannot distinguish between the legacy meaning and the ClickUp meaning, which is why behavior-based sourcing (agent count, employer, tooling) works better than title search here.
What is the actual signal to filter on if not the title?
The agent count. Andy Cabasso publicly oversees 37 agents. Anyone claiming responsibility for more than 10 is operating at the top decile of current practice, per KPMG's Q1 2026 finding that only 9% of businesses orchestrate multiple agents across workflows. Look for that number in About sections, personal sites, conference talks, and GitHub READMEs, not in job titles, because the title vocabulary has not caught up to the practice.
Who else is competing for the ClickUp 290 right now?
The obvious buyers are GitLab (which restructured for the agentic era the same week), Salesforce Agentforce, ServiceNow, Notion, Asana, Monday.com, and Relay.app. The less obvious buyer is Polsia, Ben Cera's one-person startup that raised $30M at a $250M valuation and is hiring exactly this profile. Series A founders should move fastest, because the big-co recruiters are still building their boolean strings while the pool is liquid.