That’s the policy Zeb Evans, CEO of $4 billion all-in-one productivity platform ClickUp, put in place earlier this year (before Mark Zuckerberg announced his own AI CEO clone ambitions). Evans admits it was jarring for his 1,300 employees, but it was all part of his strategy to force them to get comfortable defaulting to AI agents as he began rewiring the entire company—both its internal operations and product offerings—around them.

After going all in on agents in January, ClickUp now has around 3,000 internal AI agents embedded directly into workflows across departments, making for a roughly 3:1 ratio of agents to employees. Essentially every workflow and task now involves agents, which employees direct to autonomously plan and execute complex series of tasks on their behalf. It’s a fundamental change in what work looks like for each employee, as well as how the company does business as a whole. 

“The biggest shift is from actually doing and waiting on the work, to reviewing the work and ensuring that it meets your standards,” Evans told Fortune. 

ClickUp is not alone in starting to integrate agents in a deep, fundamental way, but the company is pushing the frontier of AI faster than most—including selling the agentic tech it’s now using internally as a product to customers. While 22% of businesses are exploring agents and 14% are deploying them, only 9% say they’re orchestrating multiple AI agents across workflows, according to KPMG’s Q1 2026 Global AI Pulse report.

Specific functions like software engineering have shown promising results, but such widespread use of agents is still far from proven. One 2026 study by researchers at AI training firm Mercor evaluated agents from top-tier models on 480 workplace tasks and found that every agent failed to complete most of its duties. No one—ClickUp included—has it all figured out, but the company’s headfirst dive into the world of agents provides an interesting peek into how an embrace of agents is currently playing out. 

The new agentic workflows

ClickUp employees used to spend their time on typical business tasks—drafting documents, evaluating data, reading emails. Now agents carry out the work, while employees spend their time directing those agents on what to do, offering feedback, and reviewing their work. Employees now largely act as managers of agents, both their own personal AI assistants and task-based agents that are shared throughout the company. 

Evans, for example, no longer checks email, company chat threads, or project dashboards. An agent scours it all on his behalf, decides what he needs to know, formats the essential information to look like a newspaper, and chats it to him.

“Most of my sidebar here is chats with agents,” he said, giving a tour through his digital workspace. “I’ve got a few chats with humans, but the vast majority of them are with agents.”

Arianna Young, ClickUp’s principal of demand marketing, described how she created an agent to help her scale up the company’s webinar program after noticing it was really taking off and generating significant leads. Named Wall-E, the agent now coordinates speakers, scheduling, and execution—“those paper cut activities that really add up”—allowing her to scale from one webinar per month to six.  

“It’s able to do a lot of that manual task creation and coordination that used to be on my plate, and that was really a big barrier to scaling some of these programs,” she said.

Challenges and change management

Such a drastic change to how a company operates naturally comes with challenges. One of the biggest, according to both Evans and Young, is making sure the agents have enough context to properly execute. (This is also the problem the company positions its product as a solution to.)

For example, the webinar agent once treated confirming a moderator like a top-tier red alert, going off and tagging in (human) teammates, because Young had expressed that it was important. She had to instruct it to apologize to everyone for the false urgency.

“From a human perspective, we’re not always thinking about specifically how we’re saying things. And so being able to learn how to speak to the agents in a way that is extremely clear and doesn’t have some of that natural human hyperbole is a little bit of a different shift,” she said. 

Guardrails are also absolutely crucial: ClickUp agents cannot delete anything or merge code to production, Evans said. Such a large fleet of agents also poses the challenge of keeping track of them—and the costs of running them. ClickUp has an agent org chart, which lists all the agents by name and who owns each of them. That person is responsible for making sure the agent works as intended, and when the agent needs information, it messages that person. Scrolling through the dashboard, Evans explained how it also shows exactly how much each agent costs to run (and discovered one that costs $9 every time it runs; something to keep an eye on, but not an immediate concern because it doesn’t run too often).

There’s also the change management aspect of it. He feels this change has to come from the top, but leaders need to be in the weeds with the technology themselves, to provide explanations for the changes they push, and return benefits to the employees. When an employee “10Xs a workflow,” Evans said, ClickUp increases their compensation as a direct result. (When asked for specifics and how many times this has happened so far, a ClickUp representative said the company has issued monthly performance awards that include equity grants since 2022, averaging around five per month. In the past year, criteria was shifted to explicitly prioritize employees who leverage AI to drive outsize business outcomes.)

“I think that’s really important to do because it can be scary thinking, ‘This is going to replace me,’ versus seeing: ‘This is actually going to help me not only do my job better, but it’s also going to help me get rewarded for doing it,’” he said.