Qovery has added multiple artificial intelligence (AI) agents to its DevOps automation platform that, in addition to responding to natural language prompts and executing complex operations, also anticipates the next step in a workflow.

Company CEO Romaric Philogène said the AI DevOps Copilot family of AI agents automate a range of tasks, including provisioning environments, optimizing continuous integration/continuous delivery (CI/CD) pipelines, observing processes, enforcing DevSecOps governance policies and identifying ways to apply best FinOps practices to rein in costs.

Trained using data collected from more than 25 million applications and 30 million infrastructure operations over the last five years, the AI DevOps Copilot agents automate tedious tasks that are usually at the heart of any existing bottleneck in a DevOps workflow, said Philogène.

Built using the Claude large language models (LLMs) from Anthropic, Qovery also expects over time to enable DevOps teams to also invoke LLMs from multiple providers, said Philogène. For example, a team can request simple instructions that unused environments be shut down at the end of the day or that a new service be deployed at a specific time.

Additionally, DevOps teams can automatically trigger daily integration tests to run only if the platform has been found to be unstable for more than a certain amount of time.

All actions, however, are also bound by role-based permissions, requiring explicit approval to perform tasks such as deleting databases or applications. Just as importantly, private data and credentials are never shared with an LLM.

The Qovery platform itself is designed to support multiple tools rather than requiring organizations to standardize on an automation framework that only supports a narrow range of options, noted Philogène.

It’s not clear how widely DevOps teams have adopted automation platforms but many of them are still spending too much time working on low-level plumbing issues, said Philogène. The goal should be to leverage AI and automation to eliminate as much of manual toil as possible, he added.

Hopefully, as more organizations embrace platform engineering to better manage DevOps workflows at scale, the number of teams that are provided with access to an automation platform will continue to increase. It may not be possible to automate every task, but one of the core tenets of DevOps has always been a ruthless commitment to automation.

Unfortunately, over the years DevOps pipelines have become fairly brittle, mainly because of a dependency on scripts that are all too often challenging to maintain. More challenging still, many of the pipelines that have been constructed are often redundant simply because it’s too difficult for multiple teams to reuse them.

Like it or not, the rise of AI coding tools for better or worse is likely to force the DevOps automation issue. The simple truth of the matter is the volume of code being generated is inevitably going to overwhelm existing pipelines and workflows. The issue then becomes to what degree can DevOps teams rely on AI to modernize workflows in a way that enables them to build more applications than ever in the same or less amount of time they do now.