The AI Paradox: Why Faster Coding Isn't Solving Your Productivity Problem

Yeah, I- I'm, I'm really excited because, you know, I, I think for us, going and meeting a lot of companies and working with them to improve their operating models through the modern product work that we do, um, there-- W- we have to so much, even with really large company, rely on vibes of how engineering is going, or, you know, at worst-

Mm-hmm

... it's a little bit of velocity, maybe it's a little bit of commitment, reliability or, say, due ratio.

You know, on occasion, we get a, a DORA dashboard that, you know, hasn't-- no one's done bug hygiene reporting in five years, and so, you know, you, you know, the average bug timeline is like two hundred and thirty days.

Uh, and so, um, yeah, I, I was excited when, when, when we started meeting to be able to, to be able to work with someone that can actually get some telemetry on, on the underlying workflow.

And I, I think really cool now, especially because, you know, you guys kinda get to see the matrix, right? You got thousands of engineers-

Mm-hmm

... and get to see the I, I think the answer to the question that I'm hearing a lot from not just CTOs but CEOs, which is, you know, what is the impact that AI is having and what does the data say about what AI is doing to teams? I mean, you know, there's data out there on, you know, from GitHub and different people that are seeing the amount of, you know, Claude-generated code or that, that type of thing.

Mm-hmm.

I guess generally, you know, since this whole AI boom has, has taken off, and I know you guys, you know, dogfood pretty heavily as well in terms of using it yourself and putting the telemetry on, on your own developers, what are some of the, you know, trends that people might expect and what's kinda surprising? Because I think there's a lot of noise out there of the hype and anti-hype of what is the impact AI is actually having on the, the development life cycle.

Yeah, it's a super interesting question because there's so many levels and stages to it.

Yeah.

Right.

I think, um, Dora put out a, a paper last year, late last year that, that we're seeing still holds pretty true, which is that somewhere around eighty-two percent of individuals using AI, uh, specifically in the SDLC, felt significant productivity increases.

And I think the number was like sixteen, only sixteen percent of organizations felt those same increases.

And if you think about that, that generally and then specifically on the SDLC, it comes down to the fact that a lot of the work we're doing is, is task-oriented, and the tasks are, are, you know, growing.

Don't get me wrong, right? We're not just doing code review or just generating code or whatever, but AI is inherently like a single player exercise.

And a lot of the processes that we've built over the years with, with engineering have been set around teams and team performance and team topologies and team structure and all that.

And, and AI is not really built around that, that concept.

And so you have this like inherent challenge.

You know, it's funny if you use Claude, like it, it almost feels like we're twenty years in the past emailing documents around now because-

Hmm.

Yeah

... it's just really hard to collaborate.

I had someone, a developer in a workshop last week say that anytime they open up a fifty-page markdown file from a product person, their eyes bleed, is what they said.