In a panel discussion recorded on May 8, 2026, three engineering leaders react to the 2026 State of Engineering Management Report. They note a widening AI‑enablement gap: the top 10% of companies now author about 20% of pull requests (PRs) with autonomous agents, while the median is only ~2.5%. The panel calls this the “agentic barrier,” a scaling challenge that could leave lagging firms behind. New large‑language models released at the end of 2025 are described as a major quality jump, expected to boost agent adoption and productivity further. Participants debate what faster code generation means for feature delivery, emphasizing that raw PR velocity is less important than end‑to‑end workflow improvements. Concerns are raised about measuring AI ROI: CFOs are demanding visibility into token consumption and cost‑effectiveness, and many engineers feel they need better data to assess impact. The conversation also touches on market sentiment, debunking the “SaaSpocalypse” hype and highlighting new AI‑driven product opportunities. Finally, leaders report that AI tools free engineers from grunt work, leading to more excitement, roadmap focus, and in many cases, team growth rather than shrinkage.
Chapters
It makes a lot of sense to me that companies aren't able to do AI enablement across their entire engineering org as quickly as things are changing. Like, we noted that our SEMR a year ago didn't even have clawed code on it and now it's the top coding tool that people are using.
We're seeing this kind of, this widening gap. The 90th percentile, top 10% of companies, they're authoring 20% of their PRs with autonomous agents and the median is like 2.5%. The ability to scale this is a real barrier. You know, we talk about it as like the agentic barrier and, you know, it's hard for people to get past that, but if they don't, they start to lag and fall behind.
This is a really interesting moment where, um, newer models that dropped at the end of 2025 really were an amazing step up in terms of their quality. And I bet we're going to see this number go up significantly this year as people adopt the new models and also put some best practices around them.
Obviously you can write code faster, but like, what does that mean for how fast you actually deliver a feature? And is that moving faster? So it's really really cool to hear people saying, we do think we are getting 25% or more faster at actually delivering.
It's almost irrelevant whether it's 25 or 50 or 100% faster in terms of just how many PRs you push if you can't do all that other stuff. That's how you get to 2x, 3x, 10x, whatever you, you want to consider the, the kind of promise of, you know, a truly agent-powered way of, of, you know, building with AI.
Uh, there's a lot of concern, I think, in the software space right now from investors and in the market, kind of the uncertainty of what's going to happen to software with the introduction of AI. As much as there's concerns for maybe traditional models of software and SaaS, there's new opportunity with AI. We can build new capabilities that are, that are more powerful and more interesting. Uh, what I see when I look at this number is a lot of optimism that as much as Software's changing. It's certainly not going away. It's just maybe getting more interesting.
I guess the SaaSpocalypse is maybe overblown, right? The idea that we're all going to want to build all our own tools is just crazy. Anyone who's ever worked in engineering knows there's just way more things that you could do, way more customers you could serve, way more software you could build than is ever, you know, really possible.
By and large, most people on my team are now much more excited about what we can do with it because they are able to see They're able to see themselves using them to do the like more grunt work so they have time for more roadmap work, something we talk about in the report. And there's so many engineering leaders who are now saying, I'm not shrinking my team, I'm growing my team.
I mean, I would read it as 60% of people say they need better data and 40% of people don't realize that they need better data yet. You know, the world we're in this year is people are worried about costs, they're worried about ROI, and those are complicated things to measure. You know, 2026 is the year where the CFO is starting to ask, where are all the tokens going and what are we getting for them? So, you know, you need sophisticated ways to measure to answer those questions, right? These are, these are real resources with real consequences.
This result was zero surprise to me, maybe even surprising that the number isn't higher, I'm seeing just an incredible interest in getting good telemetry for AI practices so that leaders can then replicate their— the best practices in their org and get those outsized impacts across everything they're doing.
