A new report on software engineering in 2026 just surfaced a gap that a lot of leadership decks have been quietly papering over. Chainguard's Engineering Reality Report says 89% of organizations report their engineers save at least three hours a week using AI. That's a real number, and it's the kind of stat that ends up on a board slide. But ask the engineers themselves and the story gets more complicated: only 46% say they feel supported using AI at work, even though 55% of their leaders are actively encouraging them to use it.
That gap between what leadership believes is working and what engineers actually experience is the real story here, not the productivity number.
Where the time actually goes
The report found 72% of engineers say they don't have enough time to build new features, the part of the job that pulls most people into engineering in the first place. 93% still find coding and feature work rewarding. But on average, they spend just 16% of their week doing it.
Maintenance eats the rest. 79% of engineers name it as a major time drain: patching, upgrading, chasing down vulnerabilities, keeping the lights on. None of that shows up as AI adoption in a survey, and none of it goes away because a team bought a batch of Copilot licenses.
Automation works, when it's actually integrated
Here's the part that should change how teams think about rollout. Among engineers with high automation adoption, 94% report spending most of their time on the work that energizes them. Among engineers with low adoption, that number is 67%. That's not a small gap, and it says the tooling itself isn't the bottleneck. Integration is.
Tool fragmentation backs this up. 70% of engineers now use five or more tools a week, and 88% say switching between them costs real productivity. 44% report it disrupts focus badly enough to notice. Add an AI coding assistant, a separate agent for tickets, another for docs, and a fourth for reviews, and you've handed your team four more context switches a day, not three fewer hours of maintenance.
What we'd tell a team rolling this out
Don't measure AI adoption by how many tools got purchased. Measure it by how many fewer tabs an engineer has open by Friday. If a new tool doesn't remove a context switch somewhere else in the stack, it's not adoption, it's addition.
The organizations getting real time back are the ones that picked a smaller number of tools and wired them into the actual workflow, ticket to PR to deploy, instead of bolting AI onto every step separately. That's usually the cheaper option too, since one integrated toolchain beats five subscriptions nobody fully uses.
The leadership-engineer gap closes the same way. Encouraging AI use isn't support. Support looks like clearing enough of the maintenance backlog eating the 84% of the week that isn't feature work, picking one integrated toolchain instead of five competing ones, and asking engineers directly what's actually slowing them down before buying the next license.
We're here to help founders and teams design and build digital products that are built to scale with you, not slow you down. If you're looking to build something, get in contact with us today!