Back to home
Revenue Agent: designing an AI agent people adopt
Revenue Agent was Outreach's first agentic experience. It finds accounts and prospects, drafts personalized outreach, and sequences it, the work a rep would do, at scale.
TL;DR
The agent worked. Getting people to use it was the design problem.
Outreach's first AI agent worked, but adoption was poor: the experience felt complex, with features piled on over time. I took it over and redesigned it for simplicity, splitting it into four clear steps (Identify, Engage, Limit, Review), moving detail behind info icons, and explaining every control in plain language. The RevOps team's immediate feedback was that it is now far easier to demo, and the experience is moving into Agent Studio next.
Problem
It worked, but no one wanted to use it.
Adoption was poor. The experience felt complex and hard to use, because features had piled up over time without anyone rethinking how they fit together.
Process
Redesign it around one goal: simple enough to adopt.
Give it a shape.
Split the experience into four clear steps, Identify, Engage, Limit, Review, so a pile of features became a flow.
Quiet the noise.
Cut the visual clutter, and move the heavy explanations behind info icons, there when you need them, gone when you do not.
Speak plainly.
Explain each control in real terms (set this limit, schedule this many prospects), and let people reuse lists they already have instead of building targeting from scratch.
Solution
A calmer agent with a clear flow.
Plain-language controls, detail on demand, and nothing on screen you do not need yet. Simple enough to demo, and to trust.
Impact
Easier to demo.
The immediate word from the RevOps team after the redesign, and customers are more excited about the experience. Next, this experience moves into Agent Studio, where Outreach customers use pre-built agents, and one day build their own.
Impact
$5M in net-new ARR from AI features in under a year. Revenue Agent was one of the Amplify features behind it.
It was Outreach's first AI agent. I later led rebuilding its workflow into Agent Studio, so it became a foundation for building agents, not a standalone feature.
Next case study
Visual design