Rolling Out Agents: The People Problem Bigger Than the Tech
The hardest part of deploying an agent isn't building it, it's getting people to trust and use it. Change management is the real work.
Here's what the AI vendors don't dwell on: building the agent is the easy bit. Getting people to trust it, use it, and not quietly sabotage it is the hard bit. The technology is rarely what sinks an agent rollout, the people problem is, and change management is the real work.
Address the fear honestly
When an agent automates part of a job, people wonder if they're next. Ducking that question breeds resistance. Address it straight, what the agent does, what it doesn't, what it means for their role, because people resist what they're afraid of and afraid of what's hidden.
Bring the users in early
An agent designed in isolation and dropped on people gets rejected on principle. Involve the people who'll work alongside it in the design, so it's something they helped shape, not something done to them.
Show, don't mandate
Trust in an agent is earned by seeing it work, not by being told to use it. Let people watch it handle real cases well before you rely on it. Demonstrated reliability wins adoption; mandates win malicious compliance.
Real scenario: a client built a genuinely good agent and rolled it out top-down with no consultation. The team, fearing for their jobs, found every reason not to trust it and it nearly failed, despite working fine. We restarted the rollout with honest conversation, early involvement, and demonstrated wins. Adoption followed. The tech was never the problem, the people were, so we did the people work. Change management is the job.