Fin (formerly Intercom) CTO Darragh Curran set a public goal to double engineering productivity—and nearly tripled it. In the first episode of Leading With Observability, he talks with Charity Majors about AI-driven PR review, hands-on leadership through the transition, and why observability is the trust mechanism that makes it all work.
By: Rox Williams
How Honeycomb Helped Intercom Transform the Way They Observe and Operate Fin.ai
Watch Now A few months ago, Darragh Curran, CTO at Fin (formerly Intercom) set a public goal to double engineering productivity and nearly tripled it instead. They did so by pulling a few levers: AI writing code at scale, building an AI-driven PR review system, leveraging observability as a trust mechanism, and with leadership becoming more hands-on through the transition.
Charity wanted to pick Darragh's brain on the messy bits, not just the highlight reel, so she invited him to participate in our first episode of Leading With Observability, a digestible video series for us folks with the attention span of a goldfish—think 15-minute (ish) webinars.
Watch the interview #
Don't have 16 minutes? Keep reading below for a short recap on the conversation.
Do a thing. Learn. Do the next thing. #
For those of you who missed it, Darragh contributed to a chapter in Observability Engineering (grab a free copy of the book today) on what observability looks like from a leadership perspective. In his section, A Letter From a CTO, he gives advice to engineering teams on how to attain excellence, and it ends with a rather simple loop: “Do a thing, learn something, do the next thing. That is a core function of a software team. Pair it with vision and product judgment, and you can build great products fast. That simple loop is how I think about our jobs. Understand a problem, solve it quickly, figure out did you actually solve it, rinse and repeat,” said Darragh. “Hopefully most companies are similar in some respect, like that core framework of ‘What do we care about and how do we improve it?’ What's maybe a little bit different is just how much that's been entrenched in our DNA and trying to, as much as possible, weed out all of the things that get in the way of ‘Try a thing, learn a thing,’” he continued.
2x? No, closer to 3x. #
Back in April, Darragh made a public commitment to increase R&D productivity by 2x. When Charity asked him what led him to making such a public statement, Darragh explained that it reinforced to the team that he was serious. He also wanted to provoke more dialogue in the industry and inspire others.
“The world is different,” Darragh said. “The tools we have and how we approach our work is hopefully evolving to that new world. My rough mental model is continual doubling. We all know what happens when you keep doing that. We should be thinking of our ceiling as 100x what it is today, not 3 or 4x.”
Where observability fits #
Fin went from humans reviewing every PR to a meaningful share shipping with no human reviewer. Their approach was to mine and harvest the best feedback from their best people on each dimension and to build a system that could reliably do that. To quote Kesha’s blog post, he says, “A human reviewer typically focuses on the actual code changes, the diff. Our agent goes deeper. It traces execution paths, following the implications of a change through the codebase. This is something humans rarely had time to do, even when they wanted to.” It's not just faster, it's safer—and any human can pull the cord at any time to trigger human review.
As Darragh explained, “You've got unbounded capacity. You're not human-scale limited. There's brilliant things that humans bring to the picture, but our mental model specifically for review was, ‘How can we bring the best parts of our best people to every review?’ [...] And what if you had all of those people on their best day with infinite patience looking at your code? Obviously you could never do that before. You'd get nothing done. But you can do it now.”
Darragh continued, “Kesha and others have gone into a bunch of depth [on observability], but at a high level, measurement was so core to the approach. We had this intuition that if the review quality is below some threshold, people will tune out of it. So we had a high bar we needed to hit. We didn't wanna lose trust in the start. On every dimension then observability is important to help you spot the places you got it wrong and continue to tune the system right down to, ‘Hey, this is too slow or too expensive,’ or what happens when it goes wrong and all of a sudden people's workflow is interrupted.”
Get your free copy and learn the foundations of observability,
right from the experts.