Short answer: Adir Ben-Yehuda says coding velocity is solved but PMs and designers remain the handoff bottleneck. Autonomy AI, launched six months ago with over 160 companies, learns an existing code base so non-technical people can go all the way to a pull request.
Watch this answer from Episode 43 (starts at 1:54) · Episode 43 notes and transcript
From Episode 43: The AI Product Team Replacement? Inside Autonomy AI’s Live Demo, starting at 1:54.
Adir Ben-Yehuda [1:54]: Sure thing, guys. Thank you for having me here. It's great. My name is Adir, as you mentioned. I'm the CEO and cofounder of autonomy.ai. I've been in the industry for the past fifteen plus years with the background in go to market. And in the last two years, founded and running Autonomy AI. And, it's it's always a crazy week in AI, isn't it?
Sam Nadler [2:12]: It is. Absolutely. And I'm act actually really excited to see what you're building because to be honest, I don't really know what it is. I understand it's this organizational layer for, like, enterprise teams to begin using AI in, like, a in an organized systematic way. But I would just love to hear how does it work? What does it do? I think you're gonna show us something here, so I'm really excited.
Adir Ben-Yehuda [2:33]: That's awesome. Yeah. So let's start from the basics. Right? I think, like, generally and we're gonna talk later on about the news, but in general, the part of coding, the engineering velocity, I should say, has been resolved from that perspective, right, with AI. Engineers are no longer writing code, and they can do things very quickly and very fast. But the main problem for organizations is still the handoff. Right? How do you actually build an operating model that will support AI in general? And the main bottleneck across the board the main bottleneck across the board is that PMs, designers are still working in silos. PMs and designers are still the bottleneck between the handoff between two engineering. And even if you go ahead and say, hey, PMs can use Claude Code to actually be part of this process. What you see across the board and like many organizations are seeing the same thing, it's very hard for them. Right? They're not technical individuals.
Adir Ben-Yehuda [3:28]: They don't understand CLI. They don't understand the code base. Even if they contain the beast of Codex for that matter, they still need to be very prescribed about what they ask from that perspective. And what we've built in the last couple of years and launched it six months ago today with over a 160 companies and growing every week is this operating system on the existing code base. Our ability to actually understand an existing code base, it doesn't matter if it's like fresh new or legacy code base to understand what are the coding standards, and I can show you this in a second, what are effectively the tribal knowledge. What is legacy? What is not legacy? But also understand semantically what the code base is doing. And then with a simple web app application, you can do everything you want on your platform. No need of engineering. You can be a product person. You could be a designer all the way to a pull request and push it into production. And that's where autonomy comes in and that's the layer we've built. And in the last six months, we're just like seeing more and more demand out of this from companies that are looking to do this first or companies that have tried different solutions, right, and like failed. I like to think about us as like if Claude Code or Codex are MS DOS for the younger audience, MS DOS was an operating system back in the days.
Adir Ben-Yehuda [4:45]: It was like I gave this like metaphor to someone who was like, what? And so I'm like MS DOS was the early operating system, green line, you need to put a floppy disk, which was like how you operate a computer. We're building Windows. Right? Easy, clickable interface, you don't need to think about it. So that's kind of like the introduction to what we do. If you guys don't mind, I can jump in and kind of show you the main interface and we could take it from there.
Jordan Metzner [5:12]: Yeah. Let's do it. Perfect.
