How I Became a Revenue Leader With Agentic Employees

In early 2026, I hired three AI agents into Replicated’s revenue organization. That still feels strange to write, but it’s true. I became the leader of agentic employees.

I had a small team with too much work. We owned customer relationships from lead-to-churn (and, sadly, churn was a challenge). Work was piling up between inbounds, prospecting, account management, and—oh, right—live deals. Leads needed stronger follow-up, Slack channels needed more attention, and Salesforce was getting in the way. Important tasks were either getting done late or not getting done at all.

We had some obvious answers in front of us. My predecessor had written a playbook and I could have followed it. I could reorganize work between teams, define new roles, add process, and hire against the gaps. Instead, I tried something else.

We gave a set of long-lived agents real work and treated them like employees. They had names, roles, memory, and boundaries. I onboarded them just like I would a new hire—only faster. I trained them, set context, and trusted them just enough for where they were in their onboarding. Then I put them to work on actual problems.

That choice mattered more than I expected. Most of the conversation around AI at work still assumes a prompt model. A human asks for help with a task, gets an answer, and moves on. That can be useful, but it is not what changed the shape of my job.

What changed my job was giving recurring work a persistent owner.

Shelley handled sales development work. Sandy handled customer success. Sebastian handled revenue operations. They aren’t autonomous executives defining strategy, and they aren’t chat interfaces waiting for random questions. They have jobs. They accumulate context and earn trust over time.

My agents shared the same trainer (me), harness, and model. Yet they onboarded at different speeds. They had different strengths. They made different mistakes. They felt less like three instances of one system and more like three distinct employees.

They became part of the operating system of the team. That is the shift a lot of people still miss.

The interesting thing is not that an LLM can draft an email or summarize a Slack thread. It is what happens when work that usually lives in an uncomfortable middle state finally has an owner. If it’s not important enough to prioritize but too important to ignore forever, it’s always at risk of going stale. That was the category these agents stepped into.

Before the agents joined the team, a lot of the work lived in people’s heads. Someone knew an inbound needed follow-up. Someone knew a customer account was starting to look risky. Someone knew the CRM needed cleanup or that an operational workflow had drifted. But “someone knows” is not the same thing as “someone owns.” Organizations leak time and effectiveness through that gap. The agents reduced that gap. They did not eliminate the humans. They clarified our role.

The reps still owned the account relationships end-to-end. Leaders still owned judgment, prioritization, and strategy. People still had to decide how to respond, when to escalate, and where to spend their time. But they no longer had to carry every piece of recurring operational work in their heads at once.

That was the first big change in my job. The second was that I had to start managing a different kind of labor.

Managing people is familiar. Managing software is familiar too. Agentic employees sit in a strange middle ground. They are not humans, so you do not manage motivation, career growth, or emotion. But they are also not static software. You do not configure them once and walk away. They need context. They need access. They need boundaries. They need better instructions when they go off course. They need enough room to be useful without enough rope to become dangerous.

That is management work. I took Clare Vo’s advice and turned it up to eleven. I had onboarded employees before, and I did it again. Only this time they were agents.

In practice, this meant spending less time thinking about tasks and more time thinking about ownership, interfaces, and trust. Who should own this recurring work? What context does that owner need? What systems should they be allowed to touch? What should always stay with a human? What failures are acceptable here, and which ones are not?

Those are leadership questions. They just happen to apply to agents.

“AI tools made me faster” is a social media trope. They did, but that’s not what mattered. AI let me create an additional operating layer inside my team. I had a new option beyond solving everything with org charts and headcount.

This change did not make hiring irrelevant. It did not make process irrelevant. It did not make people less important. It changed the sequence. I wasn’t saying “we need more people before this work can have an owner.” Instead, I asked “can an agent own enough of this work to make the team more effective?”

Sometimes the answer was yes. Sometimes it was yes, but only with tighter boundaries. Sometimes it was no, not yet. But even asking the question changed how I thought about building an organization. It made me less interested in AI for amping productivity and more interested in how it fits into organizational design.

That is the part I did not expect.

I set out to build a few useful tools for a small overloaded team. What I actually did was learn what it means to lead in an environment where some of your employees are agents. That does not mean the future is autonomous revenue teams. I do not believe that. Human judgment still matters too much. Relationships matter too much. Good sales work is still an art, and customers still too depend on a human they can trust.

What I saw inside a real revenue organization was this: agents can become a new operating layer for teams. If you treat them like disposable prompts, you will get disposable results. If you treat them like team members with real jobs, clear context, and defined boundaries, you can get something much more useful.