Using artificial intelligence (AI) for coding remains a leading topic of conversation among technology professionals these days, and the big stage at Meta’s LlamaCon conference in late April was no exception.
During a session with Mark Zuckerburg, Microsoft CEO Satya Nadella said that 20% to 30% of the code in his company’s internal repositories today was generated by AI tools, adding that some projects were “probably all written by software.” When he asked Zuckerburg about AI coding at Meta, Zuckerburg responded, “Our bet is sort of that in the next year probably…maybe half the development is going to be done by AI, as opposed to people, and then that will just kind of increase from there.”
It’s not just the top tech companies using AI for coding, the majority of developers use it for at least some tasks. A GitHub survey found that 92% of responding developers use AI coding tools in professional and personal settings.
Using AI for coding is quickly changing almost all aspects of software development. IBM Fellow and Distinguished Engineer Kyle Charlet, who also is CTO for IBM’s Z Software, said that AI coding has brought a dramatic shift both in terms of its effect on the industry and on the pace of change.
“What I thought was ‘cutting edge’ six months ago is no longer cutting edge at all. That’s how quickly this movement is happening,” said Charlet. “AI coding technology is changing everything that developers do and how we think about what we are doing.”
Ryan J. Salva, senior director of Developer Experiences at Google, explained that there is a spectrum of different types of AI coding with the progressive release of tools providing direct control over code as it is written. Salva previously worked at GitHub leading the team developing Copilot, and at Microsoft as a Director of Product Management for Azure DevOps.
Salva said coding with AI actually began five years ago with predictive ghost gray text in coding tools, then moved into conversational AI, where a developer asked AI questions about the code or how to improve it. Today, developers are using what’s often referred to as ‘vibe coding,’ where they start with a natural language prompt and the model makes large changes on their behalf, often in multiple files, he said. The fourth phase is semi-autonomous or autonomous agents performing large-scale tasks such as code reviews, writing internal engineering documentation, or performing migration or modernization projects at a scale where it’s impractical for humans to review every line of code.
“With vibe coding, you are still at the laptop observing the changes in a fairly tight loop. Whereas when you work with SDLC (software development life cycle) agents, they are effectively working with autonomy, and a developer might come back later and inspect the output,” said Salva.
AI tools improve the development process
However, AI coding tools affect more than just the actual task of writing product code. Chintan Mota, director of Enterprise Technology at Rizing/Wipro, has used AI tools such SAP Build, Cursor, and Replit to design Proofs of Concepts and pilots, estimation frameworks, and build prototypes. “AI is fundamentally changing how we build software—far beyond just code generation,” said Mota. “What’s exciting is how AI is now supporting the entire software development lifecycle—from drafting design specs and writing clean code, to unit testing, peer reviews, and debugging.”
For example, Wipro built a working pilot for automating quality control in an SAP environment for the manufacturing industry using AI, which spent just a few hours to accomplish this, instead of the typical two weeks using traditional coding. Moto explained that AI helped Wipro go from rough idea to functional prototype—complete with logic, testing scaffolds, and a usable front-end—without waiting on traditional dev cycles, making it easier to secure stakeholder buy-in early and iterate quickly.
Charlet pointed out that code explanations are another way developers can use AI to improve the process.
“Let’s unleash AI on an entire code base and help developers understand it,” said Charlet. “What are the dependences with one another? What are the dependencies on IO? What are the dependencies on data access like? How are these systems all pulled together to give a really nice topological view of all the interconnectedness of the space?”
Benefits and challenges of using AI tools for coding
Similar to most new technologies, AI coding tools provide benefits while also introducing new challenges. Not surprisingly, the biggest benefit of AI coding is an increase in productivity for developers, which significantly increases the ROI of using AI coding tools. According to Deloitte, with even a 5% productivity gain, generative AI coding in the U.S. could be worth $12 billion each year.
With more time available, developers can focus on tasks that only humans can perform. Charlet said developers are often asked to complete many mundane tasks during a development project. “But these aren’t the tasks that developers were trained to do. They were trained to innovate. They were trained to build systems. They were trained to win in the market. AI coding tools are really good at removing the noise and mundane tasks around these activities, so developers can focus on what they were trained for,” Charlet said.
Additionally, AI can help address the skills gap currently facing the developer community, such as IBM using AI assistant watsonx to address the developer skills gaps by translating COBOL to Java. Salva explained that AI helps bridge the language gap because they no longer have to worry about switch-case statements of the specifics of Rust vs. Closure vs. Python vs. C.
“I can paint a vision of the thing that I want to create, and work through it in terms of what’s the end goal that I’m trying to achieve, and then let AI fill in the syntax. For me, this is one of the great challenges. Every team that I have ever worked on has a long backlog and ideas just coming out of their ears,” said Salva. “The thing that is preventing them from shipping all those ideas is capacity. AI allows developers to move a little bit faster, and hopefully as we spend more time investing in the quality of output as well, faster at quality, but it’s also inviting more people to participate in the creative process. “
Using AI coding also improves the developer experience. A survey of 500 software engineering leaders and practitioners published in early 2025 found that 95% of engineering leadership believed artificial intelligence tools can reduce burnout. Charlet shared that AI coding tools also can help with onboarding and training, which improves code consistency.
“There are dozens of coding style approaches for a task, but only one way an organization wants it done to adhere to their standards and principles,” said Charlet. “AI helps new developers and even experienced developers produce code generated in the company’s specific style.”
However, some research shows that AI is writing bad code that needs to be debugged. A DevOps.com survey found that 59% of developers said AI tools are creating deployment errors in their code at least half the time. As a result, 67% spend more time debugging code generated with AI. By making sure that they understand all of the code generated by AI, developers can then more easily and quickly fix any errors.
Competitive advantage of AI
With AI coding technology moving at a rapid pace, organizations are increasingly looking at ways to add AI coding to their processes. Charlet said companies that leverage AI will emerge as the winners. He feels that these companies will be the ones that can not only respond to market demands most quickly, but also innovate and create new markets.
“The same is true of developers. Developers that really lean on AI will have a significant leg up because not all algorithms have been invented,” said Charlet. “There is so much more to do. While improving the efficiency across organizations, AI enables developers to focus on innovations that make a difference in creating new markets.”
Jennifer Goforth Gregory is a technology journalist who has covered B2B tech for over 20 years. In her spare time, she rescues homeless dachshunds.
