GenAI is changing how we code, but coding is not going away, says Google's Gabe Monroy

Gabe Monroy, Vice President of Developer Experience at Google Cloud.

Generative artificial intelligence (AI) is dramatically changing the way developers code, but coding itself isn't going away, a senior Google executive told Moneycontrol.

"I think the landscape is changing significantly. The way developers wrote code before generative AI, how they code now, and how they will in the coming years is likely to change significantly" said Gabe Monroy, vice-president of Developer Experience at Google Cloud.

Monroy said that the composition of software teams and the roles one would typically have in a software team are also likely to change over time, due to the rising adoption of coding assistants such as Google's Gemini Code Assist.

"There will be less need for specialisation, testing, DevOps practices and distinct back-end versus front-end developments. You could have fewer developers who could be more generalist, thanks to artificial intelligence providing the knowledge of adjacent areas in the software process," he said.

For example, a back-end expert could get assistance with front-end tasks from assistants such as Gemini Code Assist, he said.

Monroy noted that as generative AI makes development easier, more applications are expected to be built. "There are far more applications that need to be developed than there are developers to build them," he said.

Read: Google Cloud's Q2 revenue crosses $10 billion, operating profit $1 billion

Gemini Code Assist, previously called Duet AI for Developers, acts as an AI-powered collaborator within a developer's environment, helping developers write, review, and debug code more efficiently. It was first announced in May 2023 and was made available to all developers in December 2023.

Developers can use Gemini Code Assist across many popular IDEs (integrated development environments) such as Visual Studio Code, Cloud Workstations, Cloud Shell Editor or JetBrains IDEs such as IntelliJ, PyCharm. It competes with other AI coding assistants such as Microsoft's GitHub Copilot, Amazon's Q Developer (previously CodeWhisperer), and GitLab Duo.

Monroy said Gemini Code Assist is still "in its early days". However, the product has garnered significant interest and traction from developers, particularly due to its support for a large context window, which allows the code assistant to process up to 100,000 lines of code.

With Gemini Code Assist, Monroy claimed to have seen up to 50 percent productivity gains across the software development lifecycle, particularly in areas such as software requirements development, UI/UX design, and test automation in startups, and about 20-30 percent productivity gains in large companies.

Another key reason for the rising adoption of Gemini Code Assist is its support for code customisation, which tailors the coding assistance to private code within enterprises.

"Most developers aren't interested in using AI assistants based on publicly available techniques, but they also need AI assistance that maps to their private code base, functions, and how they do logging and testing inside their company" he said.

Read: Google woos India's booming AI developer community with new tools, access to latest models

In August, Google Cloud announced a strategic partnership with information technology (IT) services major Wipro to integrate Gemini Code Assist into its development workflows.

At the time, Anup Purohit, chief information officer at Wipro, had stated that his company is using conversational AI/Gen AI technology to improve the efficiency, effectiveness, and quality of the work done by associates on an individual basis.

"Tools like Google Cloud’s Gemini Code Assist will significantly enhance first party productivity among our developers enabling them to deliver higher quality solutions more efficiently,” he said.

Accenture has also integrated Gemini Code Assist to its flagship AI-based automation platform GenWizard to improve the speed and quality of software projects while Capgemini has been using Code Assist to boost software engineering productivity, improve code quality, security, and enhance developer experience.

Monroy mentioned: "The Indian system integrator community is one of the largest pools of developers. So, if we can make Gemini Code Assist work for that class of customers and accelerate their productivity, it will benefit people worldwide."

He emphasised that a big focus area while rolling out Gemini Code Assist among enterprises and global system integrators (GSIs) is ensuring that it addresses compliance, legal, and risk mitigation concerns. "We need to ensure that there are robust business cases, and validation, along with close arm-in-arm work with partners, so that we can slowly and thoughtfully sequence the rollout to thousands or even hundreds of thousands of developers in an organization."

Read: Airtel inks deal with Google Cloud to accelerate cloud adoption, deploy Generative AI solutions

Going forward, Monroy highlighted a growing trend toward semi-autonomous agents in software development. For example, instead of a developer simply prompting, 'Please build a function to add an item to a shopping cart,' and receiving the relevant code, one could make more complex requests, such as, 'Help me build an entire shopping cart application with lifecycle management and specific characteristics'.

"This would result in a series of automated prompts and responses from the model, orchestrated in a way that reduces developer toil and significantly accelerates productivity" he said.