Harness releases generative AI assistant to help increase developer efficiency

It’s no secret that there aren’t enough developers for the amount of work out there, so anything that helps ease their work load is going to be welcome. Since Harness launched five years ago, it’s been focused on creating a developer toolkit to help developers operate more efficiently throughout the software development lifecycle.

That has at least partly involved using machine learning models to identify areas that could be improved, and today the company announced the release of the AI Development Assistant, or AIDA for short, Harness’s generative AI assistant.

Company CEO and founder Jyoti Bansal says AIDA is an extension of a lot of the work the startup has been doing over the years. “When we look at generative AI we look at how Harness has such a proven history of bringing AI to DevOps, CI/CD and deployment, verification and all the different things we have been doing,” Bansal told TechCrunch.

Bansal sees a lot of companies looking at generating code as the key generative AI advantage, but he sees a much broader set of use cases than that, ones that can improve developer productivity as much as 30%-50%.

“The entire software development life (SDLC) cycle involves multiple phases including writing code, building code, testing code, ensuring security and ensuring reliability, deploying changes, verifying changes, ensuring the right costs — and that’s what we are looking at as we bring generative AI to all of these elements of SDLC to increase the productivity and efficiency in each of these different stages,” Bansal said.

He says that the goal is to infuse generative AI in every part of the Harness platform. While the AI assistant is a work in progress, for starters it involves three key elements. For starters, they are offering automatic resolution of build and deployment failures.

Bansal says that as developers make changes, it can have impact on the many systems a typical program touches, which could include an AWS account, the HashiCorp secrets manager, a Kubernetes cluster and so forth. He says changes can cause any one of these multiple interactions to fail, forcing the developer to track down the cause of the failure. Instead you can ask for the cause of the failure and the fix. The developer controls whether to implement the fix or not, keeping humans firmly in control of the process.

The second piece involves finding security vulnerabilities and automatically fixing them after the developer approves the fix, and finally looking at helping control cloud costs using natural language to help find savings.

The company deliberately named the new tool an AI assistant because the purpose is to help speed up the work process, rather than replace developers. The humans remain in control because Bansal says the fixes won’t necessarily always be right.

“We’re assisting the development process. We’re not taking over the development process. The developers are involved in it. They still have to do the work that they were doing before, but it could be just more efficient, perhaps making them 30%, 40% or 50% more efficient in what they were doing,” Bansal said.