Harness releases Release Orchestration features with AI-enabled validation and rollback

Even though organizations rely on AI to code and engineering teams release daily or more frequently, those same teams report that nearly a quarter of their deployments require maintenance, and their maintenance time comes in at over 7 ½ hours.

This, according to the 2026 State of DevOps Modernization report from Harness, clearly shows that the outsourcing process is causing the problem.

To address that gap, Harness is today unveiling new Release Orchestration capabilities with AI-enabled verification and rollback that the company said automatically determines whether each release should proceed. And now native to the company’s pipeline are native feature management and testing (from its acquisition of Split Software in 2024) and Snowflake’s Database DevOps, so changes to code and data go through the same delivery flow, the company said.

“What we’re seeing on the ground is the amount of code that developers and tools are doing, and the whole system can’t keep up,” Bradley Rydzweski, senior vice president at Harness, told SD Times. “So you have this big bug where you have this incredible speed of coding, and it just hits a brick wall because there just aren’t enough people to review it and release it. Basically, everything follows the code.”

The point of today’s Harness release, he said, is “just to help alleviate those pain points.”

Among the features of release orchestration is the ability to coordinate multiple releases as a single, unified and automated process in the Continuous Delivery Harness, rather than on demand. Wires and spreadsheets are still used to coordinate multi-team releases, the company said in its announcement. With new capabilities, Harness enables work to move through the delivery process with shared orchestration logic and common controls, gateways, and sequencing. This makes the releases behave more like systems than handouts, according to the announcement.

As code speeds, organizations are creating more features than ever that need to be shipped. Managers are always checking with the teams to see where the feature is.. in development, or QA or production. “You see these issues not just in terms of the delivery process and the delivery software, but as the bottleneck goes up to the leadership,” Rydzewski said. The power of the feature flag it evaluates each release using your existing data and decides whether to go forward, pause, or roll back.

On the Database DevOps front, which now supports Snowflake, it moves schema changes through pipelines and code. This, the company said, is especially important for teams building AI applications in data warehouses, where schema changes are constantly evolving and resulting. “We bring AI and Continuous Delivery to your databases, we automatically manage migrations and database changes and rollbacks, but we also provide that layer of governance,” Rydzewski said.

Rydzewski said outsourcing involves changing operations, using automation to make decisions and ensure delivery is going well, or stopping it if it’s not.

“It all comes down to modern reporting in terms of pain points,” he said. But have organizations realized the vision of AI? “I don’t think we have it yet,” Rydzewski noted. “We’re coding faster, but are we releasing faster? I don’t know that businesses are like that, so I think that’s our goal now. If we can solve this problem and unlock everything for AI after code, I think then, and after that, we can see the full benefit of AI.”