GitLab this week made generally available a version of its generative artificial intelligence (AI) framework for software development that includes AI agents developed by Amazon Web Services (AWS).
David DeSanto, chief product officer at GitLab, said GitLab Duo with Amazon Q gives organizations that are building applications on the AWS cloud the ability to create more tightly integrated DevOps workflows using both AI frameworks.
Amazon Q Developer is a set of AI agents that AWS has developed using the same framework it used to create Amazon Q Business, a set of AI agents capable of automating tasks across a wide range of business workflows.
The integration with Amazon Q Developer will enable DevOps teams to integrate a set of AI agents designed to be used primarily by application developers with a set of generative AI capabilities that GitLab is providing across the entire software development lifecycle, said DeSanto.
Collectively, that integration will make it simple to, for example, analyze requirements, plan implementations and generate merge requests. They can also be used to identify the root cause of a vulnerability, align with surfacing code to remediate it and review code.
Additionally, AI agents can also accelerate the process of updating legacy code bases to a more current version of a programming language.
It’s not clear how widely generative AI is being applied to software engineering, but a Futurum Research survey finds 41% of respondents expect generative AI tools and platforms will be used to generate, review and test code. Many developers are already clearly using these tools to generate code, but many have also found it challenging to debug code they didn’t write simply because they don’t understand how it was constructed.
Many software engineering teams are also discovering that because AI coding tools were not exposed to the unique attributes of the platforms being used to deploy the code they create, it won’t run without being reworked by a human developer.
There is little doubt that AI tools won’t be relied on to automate a range of tasks, but some care needs to be taken to understand when and how best to apply them, said DeSanto. Generative AI tools are, by definition, probabilistic in the sense that they generate output based on their ability to guess the right answer. Many DevOps workflows, however, are deterministic in that they need to be completed the same way every time, he noted.
The challenge then becomes determining which use cases make the most sense for using AI tools rather than simply applying AI for the sake of AI, said DeSanto.
Ultimately, AI tools will be a boon to software engineering in that many of the manual tasks that today make software development more tedious than anyone cares for will be eliminated. The thing that remains to be seen is how AI agents might soon take advantage of the reasoning engines built into the next generation of large language models (LLMs) to automate not just specific tasks but potentially entire workflows.
