Lead-Train-Incentivize AI Adoption Sprint
Model the behavior, teach it visibly, then reward a short burst of practice.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 93%
The Lead-Train-Incentivize AI Adoption Sprint combines role modeling, practical training, and a short competition to change team behavior. First, the leader uses the tool personally so the request carries evidence rather than executive distance. Next, the leader records screen shares or long-form demonstrations that show complete workflows and make the opportunity feel tangible. Then the organization runs a time-bound challenge tied to measurable output, such as pull requests opened and merged with an AI coding agent. A financial reward adds urgency, while a public leaderboard makes progress visible and encourages peer learning. The sprint works because it compresses experimentation into a shared event and rewards applied competence. Quality controls still matter: only verified outputs should count, and lessons should be shared after the contest.
Origin
At Gumroad, Sahil combined personal use, team-focused video demonstrations, and a May competition that split $33,000 among employees who merged more Devin pull requests than he did.
Core principles
- 01Leaders create credibility by using the new tools themselves
- 02Visible demonstrations make unfamiliar workflows concrete and exciting
- 03Short, measurable competitions concentrate learning effort
- 04People learn faster when experimentation and cross-sharing happen together
- 05Rewards should reinforce real outputs rather than passive training completion
How to run it
- 1
Lead from the front
Use the tool in your own production work and show the resulting artifacts. This turns adoption from an instruction into a behavior the team can observe.
Pro tip Choose real work rather than a polished toy demo.
Watch out A manager who never uses the tool will struggle to make the change feel credible.
- 2
Teach the full workflow
Record screen shares or live sessions that expose setup, prompting, review, failures, and iteration. Design the demonstration for the people who must apply it.
Pro tip Use a format long enough to show the messy middle, not only the successful result.
- 3
Define a bounded challenge
Set a short period and a measurable output that represents genuine tool use. Gumroad counted Devin pull requests that were both opened and merged during May.
Pro tip Make the leader's own output the benchmark when that creates a motivating target.
Watch out Do not reward raw attempts that never pass review or reach a useful outcome.
- 4
Reward verified practice
Offer a meaningful incentive and track performance publicly. The reward makes time for learning feel sanctioned rather than like extra work competing with normal duties.
Pro tip Split the reward among everyone who clears the standard to support multiple winners.
Watch out Balance the score with normal review standards so quantity does not overwhelm quality.
- 5
Share the learning
Review the results, highlight useful pull requests, and have participants exchange techniques and failures. Use the sprint to seed a continuing practice rather than a one-off spike.
Pro tip Show the leader's ranking and misses openly to normalize experimentation.
In the wild
Gumroad offered $33,000 to be split among employees who opened and merged more Devin pull requests than Sahil during May. Sahil produced 27 and finished fourth, which meant three engineers beat the leader's visible benchmark while learning the tool through real work.
→ The competition produced hands-on adoption, visible examples, and multiple employees who exceeded the leader's own level of use.
Sahil recorded screen shares, including a three-hour public session, with his own team in mind. He used the demonstrations to show what faster development could look like after technical migrations such as adopting Tailwind.
→ The videos supplied practical training while building enthusiasm for the organizational change.
Common mistakes
Mandating tools from a distance
Telling people to change without demonstrating the workflow makes the initiative feel like another managerial demand rather than a credible improvement.
Rewarding activity instead of outcomes
Counting prompts or unmerged pull requests can reward noise. Tie the contest to reviewed work that reaches an accepted state.
Running an endless contest
The urgency comes from a defined window. A permanent incentive risks gaming and turns an adoption sprint into a distorted operating metric.
Is it for you?
Best for
Leaders introducing AI workflows to teams that need motivation, examples, and a safe reason to practice quickly.
Not ideal for
Organizations without reliable output metrics or where speed contests could encourage unsafe or low-quality work.
From the episode
Announcing a brand-new podcast: “How I AI” with Claire Vo 🔥