Automation is nothing new in the world of DevOps. You can’t really have Continuous Delivery (CD) without automation. CD has been practiced for the past 15 years and has traditionally been accomplished with scripting on platforms such as Jenkins. These tried-and-true techniques will continue and are well-documented. Let’s look at what’s new in 2025 which is – you guessed it – using GenAI to improve efficiency. Beyond Continuous Delivery, GenAI can improve DevOps as a whole, with automation in all five phases.
Plan
How do you automate planning? AI can analyze transcripts from meetings with stakeholders and developers to create user stories, complete with acceptance criteria. It can even research dependencies and surface potential impacts in your existing code base.
Sprint Planning? AI can practically eliminate the need by looking at the code committed by existing team members and their current workload to determine who is best suited to take on a user story. It can estimate the level of effort based on past performance. AI can even create an architecture plan for the stories. Of course, human team members should and will review the plans, but that’s 80% of the work done.
Build
We all know by now that AI can generate code, but that is only 10-20% of a developer’s day. While there is some efficiency gain there, other parts of the build will benefit as well. First, unless you are a Lone Ranger, you will experience merge conflicts.
AI can help speed up resolution resolving conflicts based on high-level guidance. AI can write unit tests based on the code and user stories. It can create functional tests based on the intention and acceptance criteria documented in the user stories. Want to talk about shifting left? Before you ever ask AI to complete a line of your code, you will have an automated functional test waiting to validate that it performs against the acceptance criteria.
Test
Even with the dependency and impact analysis performed in the planning phase, there is a risk of regressions. As you use AI to build functional tests for your user stories, you are creating a library that can be used for regression testing.
Not all test failures are due to bugs. Sometimes an existing test will break because of an intentional change in the process. AI can scan test failures, and determine which are due to bugs and which are due to tests that need to be updated. Then AI can heal the tests and suggest fixes for the bugs.
Release
Does your team enjoy writing release notes and user documentation? I didn’t think so. It’s especially difficult in an Agile practice since you never really know if a feature will make it into production until it does.
AI is great at assembling and summarizing information. Remember those well-written User Stories created in the planning phase? AI can determine which of those made it all the way through to production and instantly create release notes. With a little setup, you can create user-specific release notes and FAQs. Better yet, add those release notes into an AI workspace and forget about the FAQ. Simply provide your users with a link to the workspace and they can ask any question they want.
Operate/Monitor
Big Data AI techniques have been used in monitoring and observability for quite a while now, and this should continue. As Agentic AI begins to take off in Business Applications however, a new kind of monitoring will become important. Sentiment analysis tools are used to rate the ‘attitude’ of chat and email to detect angry customers.
Now with Agents communicating directly with your customers, you will have to monitor your agents’ tone of voice as well. This will require a natural language-capable tool. We can expect to see new ‘hyper emo’ AI tools that are deliberately sensitive and emotionally attuned to monitor responses from other agentic AIs. Be forewarned though, you may have to send your agents to therapy if they can’t take the criticism.
While this isn’t an exhaustive list, you can see that GenAI tools will have a tremendous impact in the way code will be developed. Many of the tedious tasks in DevOps should be handled by Agents that will enable product owners, architects, and interaction designers to focus on the things that matter most, the efficacy and usability of the software they create.
