By Ken Rona, Chief AI Officer, JWX
The third priority in JWX's AI mandate, helping business staff re-engineer their own workflows, turned out to be the hardest one. Here's what didn't work, and the program that finally did.
If you're rolling out AI to a non-technical team and assuming training videos and webinars will get you there, this post is the cautionary version, and the fix that actually worked.
The last part of the mandate was to enable AI use in non-technical workflows. Our original thinking was that we'd provide a supported set of tools for non-technical users and they'd re-engineer their own work. That didn't work. Some interesting things got built, but they weren't reliable, and the AI CoE team had to do extensive rework. Adoption, in contrast to the engineering staff, was very spotty.
We dug into why. Product management wasn't a heavy user of AI, even though the CoE team used it extensively to build product requirement documents, business cases, press releases, essentially every artifact an engineer needs from product management. But those practices never made it out to the larger product management group. We tried webinars, videos, written documentation. Adoption stayed slow.
We talked with product management leadership, and their read was that the staff was too busy to invest the time in learning the tooling on their own. So we created space for them to learn instead of asking them to find it. We built the “AI Residency" (AIR): a boot-camp-style program where the entire product management staff was excused from their normal deliverables for a stretch, with one job: become AI-proficient.
Each PM picked a project they actually wanted to build and we helped them scope it into something achievable in a week. We ran eight training sessions covering AI tools, AI in design, and our internal coding tools, backed by coaches who were advanced AI users. At the end, we ran a contest to see what people had created.
The AIR was a genuine success. One participant called it the most impactful professional training of his career.
We continue on our mission with the same three priorities. At this point, most people who were going to try AI have. That's the full picture across this series: three priorities, fifteen months, a bot family that pushed engineering output up 3x, an accidental enterprise assistant now running Slack-based feature development company-wide, a three-day prototype that became a real product, and a training program that worked once we stopped asking people to find the time and started giving it to them.
That's the full picture: three priorities, sixteen months, a bot family, an accidental enterprise assistant, a three-day prototype that became a product, and a training program that changed how an entire department works. If you missed any part of the series, here are the links to all of the blog posts: