Breaking News: IBM Unveils Bob, an AI Coding Assistant for the Enterprise
IBM today announced the launch of Bob, a new AI-powered coding assistant already in use by more than 80,000 developers within the company. The tool, developed under the leadership of Neel Sundaresan, General Manager of Automation and AI at IBM Software, represents the culmination of two decades of research into developer productivity.
"Using a powerful AI for everything is like taking your Ferrari to buy milk," Sundaresan told The New Stack. "Bob is designed for the right job—accelerating enterprise development without overcomplicating simple tasks." The quote underscores IBM's focus on targeted efficiency rather than brute-force automation.
Background: Two Decades of Developer Productivity Research
Sundaresan, a founding engineer of Microsoft GitHub Copilot, has been obsessed with developer friction since the early 2000s. His first system was not a modern AI but a recommender for API calls. "30% of developer code is API calls," he said. "If you do a class dot something, you get a long list… That itself is a friction point."
The early model used search ranking, not transformers. Developers loved it, he says, because it reduced friction at a small but critical moment. That insight—that user experience matters as much as model quality—has guided Sundaresan ever since. "If the system makes a wrong recommendation that interferes with my thought process, that matters," he emphasized.
Over the years, Sundaresan's team tracked every major AI evolution: LSTM, encoder-decoder, the Transformer paper, GPT. "If you go back and look at our publications, we have publications in all of this," he said. "Every paper would say, here's the model—but the user experience was always orthogonal to the underlying AI."
Why Bob Now?
Bob arrives at a moment when AI coding assistants have become mainstream. Unlike Copilot, which targets broad code generation, Bob is tailored for enterprise workflows—prioritizing context-awareness and integration with IBM's own software stack. The 80,000 internal users signal confidence before external rollout, which IBM plans to expand later this year.
What This Means: A New Chapter for Enterprise AI Coding Tools
Bob represents a shift from generic copilots to specialized assistants. Sundaresan's argument—that developer satisfaction comes from friction removal, not model size—challenges the prevailing race for larger models. "If you get the surface wrong, you can have a better model and a worse product," he warned.
For enterprises, Bob promises lower overhead and tighter security, as it runs within IBM's own infrastructure. But the broader implication is that the market is segmenting: consumer tools like Copilot for general use, and domain-specific tools like Bob for complex, regulated environments. "Coding is an analytical task," Sundaresan said. "It's different from shopping online."
The launch also validates an underappreciated approach: start with small friction points, not grand automation. As Sundaresan put it, "You don't need a Ferrari for milk—you need a reliable tool that gets you there without distractions." Bob aims to be that tool.
What to Watch Next
- External availability: IBM plans to roll out Bob to selected enterprise customers in Q3 2025.
- Integration with Watson: Bob will leverage IBM's Watson AI platform for deeper code understanding.
- Competitive landscape: How will Bob differentiate from GitHub Copilot, Amazon CodeWhisperer, and others?
For more on Sundaresan's vision, see the Background section above.