Generative AI (genAI) adoption continues to ramp up quickly, and many organizations are already capturing real business value in key areas such as software development. Nearly 60% of organizations are using coding assistants, and developers are seeing significant benefits, ranging from improved job satisfaction to faster resolution of developer-related issues, according to a recent IDC survey.
But coding assistants and the like are just scratching the surface: The rapid evolution of agentic AI promises to raise the ceiling further on automating and optimizing the entire software development life cycle.
“Agentic AI essentially turbocharges the capabilities of LLMs [large language models], allowing them to go from existing coding assistants to taking responsibility for the execution of tasks,” says Arnal Dayaratna, research vice president, Software Development, IDC.
A common perception of AI agents is that their main benefit is the ability to automate tasks without human intervention. In software development, however, the main difference between coding assistants and agents is that agents can perform tasks based on context rather than predefined thresholds or triggers, Dayaratna explains. As a result, agents can execute more complex tasks than coding assistants. Agents can also access and use technologies outside of a code editor or a browser.
These capabilities enable a variety of new use cases for deploying agents for software development beyond code generation and testing. For example, engineers can deploy agents to make changes to a code repository or a data store or to integrate new code into existing code to automate continuous integration workflows. Agents can take on activities for modernizing or transforming existing apps, in addition to developing new ones. They can also automate entire work streams — such as deploying an update to hundreds of legacy applications in parallel in the background.
“Agentic developer technologies enable generative AI to operate across the entire software development life cycle,” says Dayaratna.
How developer skills will evolve
As AI agents take on more development tasks, the skills of in-house engineering teams will need to evolve to keep pace. But there’s a common misconception that developers will transition quickly into a strictly supervisory role, overseeing quality assurance and control of agents in production environments.
Instead, says Dayaratna, developers will need to take on two important tasks: building the agents themselves and building systems that enable a multitude of agents to collaborate.
This transition requires new skills. “Developers will have to learn how to build solutions on top of LLMs, and they will need data expertise in order to define the right data sets for agents,” says Dayaratna. “Developer skill sets will continue to expand and deepen — they won’t change completely, but they will be transformed significantly.”
In short, engineers will need to embrace the evolving concepts of software, characterized by natural-language queries and agents, versus traditional application programming and development tools.
“The modality of development is going to change to be based on natural-language interfaces,” says Dayaratna, citing GitHub Spark as a current example. “That is the future of enterprise software development.”
Together NVIDIA and Microsoft Azure enable enterprises to create, deploy, and improve autonomous AI agents and reasoning systems. The result is better decision-making, greater operational efficiency, and transformative customer experiences.
Explore real-world use cases; the evolution of agentic AI; and how Azure and NVIDIA empower scalable, secure, and intelligent solutions — faster.