Excel 2013 introduced an AI feature, FlashFill, that could learn string transformation programs from examples. Now, spreadsheets are hot again with the recent release of ExcelAgent, which brings reasoning models directly into the Excel surface. It can read and write spreadsheets, generate formulas and charts, and even self-inspect its own results to fix errors.

AI models and agents are evolving at remarkable speed, and spreadsheets provide a uniquely fertile playground to leverage their power and guide their evolution. They combine code and data, demand precision, and offer the potential for learning and creative experiences for a diverse user base.

From early AI technologies that understood user intent from input-output examples and contextual cues, to today’s reasoning models that understand natural language, we have come a long way. Yet the true breakthrough isn’t just in what AI can do alone, but in what humans and AI can do together. To harness human tacit knowledge, agents must be designed to ask the right questions at the right time. For this, we draw on principles of cooperative communication from the social sciences, such as the Gricean maxims, and leverage techniques like distinguishing inputs and sketching solutions.

AI progress today is often benchmark-driven. But , capturing open-ended, human-in-the-loop tasks. I’ll share methods to construct such benchmarks, along with approaches for evaluating human–AI collaborative workflows that include synthesizing domain-specific rubrics grounded in conversational principles and using simulated user proxies informed by personas.

As AI takes on more critical tasks, correctness will remain essential yet insufficient. We also need innovations in context engineering, knowledge management, and verification workflows to ensure results are not only accurate but also interpretable, verifiable, and aesthetically sound. These qualities become even more vital as we expand beyond productivity (where AI helps users go faster) toward learning (where AI helps users think deeper), and creativity (where neither human nor AI knows the destination, but together they explore new horizons).

In this talk, I’ll share stories behind key spreadsheet innovations, including FlashFill, FormulaPrediction, AIFunctions, TableImport, Smart Copy-Paste, and ExcelAgent, to show how this familiar grid continues to evolve as a living laboratory for AI advancement and human–AI collaboration.

Bio

Sumit Gulwani is a Distinguished Scientist at Microsoft, where he connects ideas, people, and research with real-world practice. He is the inventor of Flash Fill in Excel—now not only a productivity feature used by hundreds of millions, but also a concept featured in middle-school computing textbooks. Sumit leads the PROSE research and engineering team at Microsoft, advancing the frontiers of AI and embedding these innovations into Microsoft products—especially in the domains of software engineering and spreadsheet intelligence. A strong believer in the power of storytelling, he sponsors internal training programs at Microsoft to help technical teams communicate with greater clarity and impact. A passionate mentor, he also runs a thriving remote research fellowship program in India, nurturing young talent. His work has earned 13 research paper awards, the Max Planck-Humboldt Medal, ACM SIGPLAN Robin Milner Young Researcher Award, and Fellowships from ACM and AAIA. He received his PhD from UC Berkeley, where his dissertation was honored with the ACM SIGPLAN Outstanding Doctoral Dissertation Award. He earned his BTech from IIT Kanpur, where he was awarded the President’s Gold Medal (2000) and later the Distinguished Alumnus Award (2024). Sumit sees AI not as a tool, but as a collaborator in human learning, productivity, and creativity.

Sumit Gulwani is a computer scientist seeking connections: between ideas, between research & practice, and with people with varied roles. He is the inventor of many intent-understanding technologies including the popular Flash Fill feature in Excel. He founded the PROSE research and engineering team at Microsoft that develops APIs for program synthesis and has incorporated them into various products including Office, Visual Studio, SQL, PowerQuery, and Powershell. He has co-authored 10 award winning papers and a total of 125+ peer-reviewed papers across multiple computer science areas, delivered 50+ keynotes and invited talks at various forums, and authored 65+ patent applications. He is a recipient of the ACM SIGPLAN Robin Milner Young Researcher Award (for pioneering contributions to program synthesis and intelligent tutoring systems), ACM SIGPLAN Outstanding Doctoral Dissertation Award, and the President’s Gold Medal from IIT Kanpur.