GitLab CEO on why AI isn’t helping enterprise ship code faster
AI coding assistants are making developers more productive at writing code. But why aren’t most enterprises actually delivering more software? That’s the question Bill Staples, who took over as CEO of GitLab just over a year ago, says he keeps hearing from customers.
“They would say: we’ve invested. We’re using these coding tools. Our engineers love them, but we’re not seeing an acceleration in our innovation velocity,” Staples says he hears when he talks to. “We’re not delivering more software faster.”
In this episode of The New Stack Makers, we sat down with Staples to discuss why coding was never the real bottleneck, how GitLab’s newly GA’ed Duo Agent Platform aims to automate the full software development lifecycle (SDLC), and why context, and not just code generation, is the key to making agentic AI work in the enterprise.
The 10-to-20% problem
Staples says he spent his first 100 days as CEO talking to more than 60 customers. He was surprised to learn that even highly regulated enterprises like financial services and the public sector are going all-in on AI. But even as they go all-in on AI coding tools, software isn’t getting delivered much faster.
As Staples noted, developers spend only 10 to 20% of their day actually writing code. That translates to maybe one to two hours per day. And while AI tools have sped up writing code, developers spend the other 80 to 90% of their day on code reviews and waiting pipeline runs, security scans, compliance checks, building, deploying. Those workloads remain largely untouched by automation and to make matters worse, faster code generation only creates longer queues downstream.
“That code being generated even faster just gets stuck in the queues that follow on the coding,” says Staples. “The pipeline’s got to run. The security scans have to happen. The compliance checks need to be validated. None of that today has been accelerated with AI.”
Context is king
GitLab’s answer is its Duo Agent Platform, which recently went GA and represents what Staples calls the start of a multi-year journey to bring agentic automation to the entire software lifecycle. The platform introduces “agent flows,” which he described as multi-step orchestrations that can take a feature request from issue through merge request, handling planning, code generation, test creation and validation along the way.
A key differentiator, Staples argued, is context. While standalone AI coding tools like Cursor, Windsurf or Claude Code work with a local codebase, they typically lack visibility into the broader project: issue trackers, bug reports, epics, pipeline history, security scans and test cases. GitLab, thanks to being an all-in-one platform, can bring all of this metadata together into a knowledge graph that both humans and agents can then draw on.
“With all of the AI coding tools that we’ve talked about, they have the local codebase,” Staples said. “But the agents themselves don’t have access to the issue or the bug report or the epic that defines why this code exists.”
Platform vs. point solutions
One question a lot of enterprises have to ask themselves in this fast-moving environment is what tools to bet on.
The flood of new AI developer tool startups doesn’t worry Staples. He sees the dynamic as familiar. GitLab has always watched innovation in the open source and startup ecosystem and incorporated those patterns into its platform. With agentic AI, he argued, the case for consolidation gets stronger. Each additional AI tool creates another context silo and another vector for privacy, compliance and governance complexity.
“Honestly, it’s no different than the world before GitLab, and it’s no different than the world that GitLab has existed in for ten years,” he explained. “Because in many ways, what GitLab has done is look at the industry, look at the engineering patterns that are successful, and design those into the platform. We look at the point solutions, the best of breed solutions, whether those are open source or commercial, and then incorporate that learning into an opinionated end-to-end platform for software engineering. In so in many ways, I’m actually really excited by the innovation happening in the startup community and in the open source community — with projects like OpenClaw that explore new approaches to agentic AI — because that’s just more ideas, more exploration that ultimately helps inform our opinionated approach building software in a platform based way.”
For now, most GitLab customers are still steering agents through chat-based interactions rather than running fully autonomous workflows. But Staples sees the trajectory clearly — and he’s betting that the company that owns the full lifecycle will be the one that finally unblocks enterprise software delivery.
The post GitLab CEO on why AI isn’t helping enterprise ship code faster appeared first on The New Stack.
