Twice the pull requests. Twice the review time for each one. That is the pattern among engineers who use AI heavily, and it is the problem Hud‘s newly launched agentic rollouts are built to solve.
The numbers come from Faros AI’s 2025 vendor survey. Teams with high AI adoption merge 98% more pull requests. Their pull request review time rises 91%. Output went up. So did the time spent reviewing it.
Hud’s new capability targets that gap directly. Agentic rollouts let coding agents take a change from pull request to production, with Hud validating every step against real production behavior. The goal is to bring the speed of shipping closer to the speed of coding.
Reading the Data
The Faros AI figures describe a velocity gain that runs into a slower review stage. Engineers using AI heavily produce about twice as many pull requests. Each of those pull requests takes about twice as long to review.
A second data point covers what happens after review. Harness’s 2025 vendor survey found that only 6% of organizations describe their continuous delivery as fully automated.
Put together, the picture is clear. Agents write code fast. Review slows down. Fully automated delivery is rare. The fastest stage of engineering now feeds into the slowest, and the more code agents write, the bigger the backlog becomes. Every team that adopts AI hits this wall.
Where the Time Goes
Hud pinpoints the friction. A pull request waits for someone willing to merge it, because it is hard to trust an agent to ship without breaking production. The agent cannot see how its code runs there. Reviewers approve changes they did not have time to read. After the deploy, someone watches dashboards. A change written in minutes can wait for days. Human-driven validation stretches time to delivery.
None of those steps involves writing code. All of them involve confidence. And confidence has been built by hand, one review and one dashboard at a time.
How Hud Builds That Confidence
Hud starts with measurement. Its runtime code sensor, added with a one-line SDK, automatically learns baseline behavior at the function level in production. One line to install. It learns on its own.
From there, every change is checked at three points.
Pre-merge. The pull request is analyzed against a production impact map built on real parameters and code flows. Hud classifies its production risk level. Risky changes come with clear context that speeds review. Safe ones can merge as fast as they were written.
In flight. As the new version takes traffic, Hud compares behavior to the baseline continuously. Regressions are caught immediately. Function-level context identifies the root cause and the pull request responsible.
Post-regression. Hud’s context supports rollback, revert, and a fix to the underlying issue. The coding agent receives deep context so it does not repeat the mistake.
The outcome: teams know the production-level impact of each change before, during, and after rollout.
What Changes for the Team
The shift is from watching to knowing. Without runtime evidence, confidence comes from people: a reviewer reading the diff, an engineer watching the dashboard. With Hud, it comes from data collected at the function level. Agents and engineers look at the same evidence. Safe pull requests merge at the speed they were written. Risky ones arrive with context. Regressions are caught, traced, and reversed. The coding agent gets the details it needs to do better next time.
One Number From the Field
Riverside provides a concrete benchmark. The platform for recording and editing studio-quality audio and video ships fast with AI. Much of its backend is real-time media code that only shows its true behavior under production load.
Riverside runs every deploy through a canary with Hud’s agentic rollout and automatic rollback. Flawed deploys are detected quickly. They are reverted in about eight minutes. Usually, that happens before a customer notices.
Shilo Magen, VP of R&D at Riverside, put it this way: “Our engineers use AI agents every day, and writing code is no longer the bottleneck. The real challenge is shipping that code to production with confidence.”
“For a real-time media platform like Riverside, even a small regression can have a major impact on the user experience. With Hud, we can automatically detect regressions and roll back problematic releases in minutes. Our engineers now spend less time worrying about deployments and more time building.”
The Founder’s Read
Roee Adler, Co-Founder and CEO of Hud, connected the numbers to the product. His argument: coding and triage have agents, while the merge and rollout still run on people.
“AI already changed how code gets written. The bigger problem now is getting that code into production safely, at the speed agents write it. Every engineering leader I talk to is trying to solve this problem. Coding and triage have agents, but the merge and rollout still run on people, because it’s scary to give coding agents accountability over the stability and reliability of the production environment. That’s the missing link in the AI software factory, and it’s the problem we built Hud to solve. With agentic rollouts, agents and engineers work from the same runtime context, at the function level, so a change is merged to production based on evidence. When something does regress, it’s immediately detected and fixed, but more importantly the agent that wrote it gets a feedback loop to learn and improve.”
Hud’s direction sits in the loop itself. Every rollout teaches the coding agent how its changes behave in production. Each new change ships with less risk than the last. That feedback loop is where Hud is headed.