Ran a tournament to stop teams rebuilding the same tools
150 engineers took part; 40% of internal tools reached the top standard
Teams at Intuit build capabilities, which are internal tools and widgets meant to be reused by other teams. They lived in Figma and Storybook, built in React and TypeScript. When they aren't findable, documented or tested well enough to be picked up self-serve, the next team that needs one builds it again. I pitched an InnerSource tournament to fix that, and ran it with the AI and Data organization.
Also in this role
- Built a framework and process for engineers to submit CFPs to high-leverage conferences, reviewing content through run-throughs and editing external blog posts
- Presented new product releases internally to a global engineering organization of 7,000+
Mandates don’t work here
Nobody disagreed that duplicated tooling was worth fixing. It just never rose above whatever was already on someone’s plate. Requiring it wouldn’t have changed that, it would only have made people resent the thing I wanted them to care about.
Working to someone else’s standard
The tournament didn’t get to define what “ready” meant. Intuit already had a criteria model for capabilities, written by someone else, and the whole program had to be built on top of it rather than beside it. I met with its author to go through it until I could teach it, then trained teams on it. A published standard anyone could check themselves against was worth more than one I could have written, because nobody could argue it was the tournament’s opinion.
So I made it a tournament
It ran over a couple of months, with each team completing a task every week against that model. Points went on a running leaderboard. At the end, teams presented to a panel of judges, and the winning team took internal budget, swag, and praise routed to their managers. That last one was the part that actually moved people.
The negotiating wasn’t about design, it was about permission. I went to the leads of the AI and Data organizations for sign-off and for public shoutouts to the winning teams, and to individual managers for buy-in, so they would advertise it and let their teams take part. Participation was entirely voluntary, which meant every one of those conversations had to make the case on its own.
The real design problem was the two hours
Every participant was giving up a couple of hours a week on top of their existing workload. That’s the whole constraint. None of this works unless what you’re asking for is visibly worth more than the time it costs, so the leaderboard, the weekly cadence and the judged finale weren’t decoration. They were what made the cost legible and made progress feel like it was going somewhere.
Everything it ran on
I built the program’s website on Mailchimp, coding custom HTML templates for it and for the newsletters: the sign-up page, the schedule, the weekly assignments and the forms to submit them, the leaderboard, and every deck and recording once it happened. Mailchimp was a fairly new acquisition, and that was the point. Building on a system the company now owned was how we found out what it was actually capable of.
Slack carried the announcements and session reminders, each one linking back to the site, so there was one place that was always right. It was also where I kept teams engaged with each other. Everyone was fitting this around their normal responsibilities, and I didn’t want anyone to forget.
Points lived in spreadsheets first. Attendance counted toward them, so I recorded it, and every total was checked there before it went on the public leaderboard. A leaderboard people are competing on can’t be wrong even once.
The org spans India, California and the East Coast. I hosted the sessions in India’s time zone and presented when a US speaker couldn’t make that slot, recorded every session, and posted the recordings afterwards. A program that only works in one time zone isn’t a program, it’s a meeting.
The finale
The last day was a virtual showcase of each team’s video. I recruited the judges, their points were added to each team’s running total on the day, and the winner was announced live. Doing all of that remotely, across those time zones, was the hardest logistics of the whole program. Afterwards I helped design the swag, collected addresses, and got it sent to every participant.
What it produced
150 engineers took part, and 40% of capabilities reached the highest tier of the criteria model, meaning documented, covered, and ready for another team to pick up without asking anyone’s permission.
It ran a second time
I was asked to run another tournament shortly before I left, and participation was higher. Teams had heard about the first one and wanted in.
That’s the result I’d point at over any percentage. The strongest evidence that an optional program worked is being asked to run it again by people nobody told to ask.