By Ken Rona, Chief AI Officer, JWX
Not every part of an AI transformation goes according to plan, and this is the story of the part that didn't. What started as a simple internal search tool for developers became something we didn't set out to build: a company-wide AI assistant that lives in Slack and now touches nearly every team at JWX.
If you're rolling out an internal AI tool and scoping it narrowly to one team “for now," this is worth reading before you finalize that scope.
While we were developing our internal tooling, my lead developer, Jay, and I started to talk about making these tools available in a chatbot. Our initial idea was to create a chatbot that let developers talk to their code base. So Jay created “Digital Jeroen", or DJ, named after our founder, Jeroen Wijering. DJ is an internal search tool that has access to some of our enterprise data stores (JIRA, Slack, Notion, Google Docs, Datadog) and all of our tooling. You can now ask DJ, in Slack, to write a user story, create a JIRA ticket based on the story, write the feature, test it, and create a merge request for a developer to review. It has been a transformative tool. DJ appends its status to Slack and it has proven to be a great tool for coordinating development within a team.
And then we had the idea of letting DJ participate in Slack conversations. On our channels, a user can invoke DJ using the @dj tag, or if it sees that a question was asked and no one has answered it, DJ will respond with the answer. On our video calls, you can say “Hey, DJ" and then ask it a question, and it responds in chat once it retrieves an answer.
DJ now forms the backbone of our internal AI tools offering and is shaping up to be a very capable assistant used broadly throughout the organization. Non-developers use it to answer questions like:
It's a bit of a Trojan horse. Once someone in sales or marketing starts using DJ for one thing, they want automation for the next thing too. It's been a great way to introduce AI to non-developers, while also making the developers who built it more productive.
There was an additional benefit we didn't expect: developers started using DJ to develop features inside of Slack itself. They'd have DJ create a JIRA ticket, write the code, test the feature, review the code, and open a merge request, all without writing a line of code themselves. Each developer's instance of DJ, and the developers, could see what everyone else was working on. They could coordinate and solve problems in the open. That led to a fairly dramatic increase in productivity.
I would be untruthful if I said the plan was to build a very capable enterprise assistant from day one. But part of any transformation effort is recognizing an adoption opportunity when you see one, even when it wasn't the plan.
DJ proved something worth sitting with: the biggest adoption win in this whole effort wasn't planned, it was noticed.
Next up: “We Built a DSP Integration in 3 Days" blog is coming up on the 27th October.