Champion-Led Agent Adoption
Roll out new AI tooling by letting a few excited early adopters pave the way for the team.
- Difficulty
- Easy
- Time to result
- ~weeks to results
- Steps
- 3
- Confidence
- 82%
A pattern for driving organizational adoption of a novel, different-feeling product: let a small group of enthusiastic early adopters do the upfront investment and produce visible results, which pulls in the rest of the team — who join to a system that already knows their codebase.
Origin
Scott Wu describing the adoption pattern Cognition repeatedly sees inside customer engineering orgs.
Core principles
- 01Novel tools with a different product experience need champions, not mandates
- 02Visible output ('PRs shipped') is more persuasive than pitching
- 03Accumulated context lowers the switching cost for later adopters
- 04Early adopters de-risk and pave the path for everyone else
How to run it
- 1
Let the excited few go first
Identify the handful of people genuinely eager to try the new thing and let them invest — giving the agent the repos, teaching it lint/CI, and running the initial foothold tasks.
Pro tip Don't force reluctant engineers early; enthusiasm carries the setup cost.
- 2
Make the results visible
As champions ship PRs and knock out work, the rest of the team sees the 'new engineer' producing real output, which builds organic pull.
- 3
Onboard latecomers into a warm system
When others join, the agent already knows a good amount about the repositories they work in, so they get value almost immediately.
Pro tip Accept that this is a genuinely different product experience — invest in making the steps clear to newcomers.
Watch out Expecting instant company-wide uptake without champions underestimates how different the workflow feels.
In the wild
A few excited engineers set up Devin and gave it foothold tasks; over time the rest of the team saw 'this Devon person that just joined the company is just knocking out PRs' and got their own accounts — by which point Devin already knew the repos they worked in.
→ Bottom-up adoption spreads across the engineering org without a top-down mandate.
Common mistakes
Mandating before proving
Pushing a novel, different-feeling tool on the whole team at once, before champions have generated visible wins, meets resistance.
Is it for you?
Best for
Engineering leaders and champions introducing autonomous agents or other novel tooling into an established org
Not ideal for
Tiny teams where everyone must adopt at once, or tools with no accumulating-context advantage
From the transcript
“there will be a few folks at the team who are really excited uh and want to try out the new thing”
“the early adopters themselves can really pave the way I think for for everyone else on the team”
“by the time they join Devon already knows a good amount of detail about the repositories”
From the episode
Inside Devin: The world’s first autonomous AI engineer that's set to write 50% of its company’s code by end of year
Scott Wu (CEO and co-founder of Cognition)