Asha Sharma

Asha Sharma is CEO of Xbox, following her appointment as Executive Vice President and CEO of Microsoft Gaming on February 20, 2026; her title became CEO XBOX effective April 23. Previously Microsoft’s President, CoreAI Product, she presented Azure AI Foundry’s agent-platform agenda and, with Rolf Harms, argued that AI products should improve through usage feedback, fine-tuning, and evaluation. In her gaming leadership message, she committed to protecting human creative work as AI becomes part of game development.[2][3][4]

In her February 2026 message, Sharma set three priorities: great games, a renewed commitment to Xbox beginning with console, and new ways to play. She connected her promotion of Matt Booty to the first priority, emphasizing his understanding of game development and developers’ trust in him. Booty became Executive Vice President and Chief Content Officer, reporting to her. She also proposed shared tools and a platform through which developers and players could create and share their own stories.[2]

Sharma’s stated position on AI paired technological innovation with human authorship: “As monetization and AI evolve and influence this future, we will not chase short-term efficiency or flood our ecosystem with soulless AI slop.” She described games as art crafted by humans and supported by innovative technology. This was a commitment about how she intended to lead the business, rather than a rejection of all AI use.[2]

Before moving into gaming leadership, Sharma’s public work centered on infrastructure for building and operating AI applications. As President, CoreAI Product, she authored the May 19, 2025 Azure AI Foundry announcement, framing more capable agents as requiring models, integrated tools, and built-in governance. She described connected agents that call one another as tools, alongside stateful workflows that manage context, errors, and long-running processes such as financial approvals. Microsoft was also unifying Semantic Kernel and AutoGen to support orchestration.[3]

The announcement explained agentic retrieval in Azure AI Search, then in public preview: conversation context and an embedded language model divide a question into subqueries, run searches in parallel, and assemble an answer with citations. Sharma also announced previews of observability features for tracking agent performance and tool calls, and described evaluators that check whether agents follow user intent and use tools correctly. These capabilities addressed the operational demands of deploying agents beyond a demonstration.[3]

Sharma presented Foundry Local as a Windows and Mac runtime for applications that work offline, retain sensitive data locally, and reduce bandwidth costs, including factory-floor and remote field-service scenarios. She also introduced experimental research work: Project Amelie, powered by RD Agent from Microsoft Research, could take a task such as predicting customer churn and build a machine-learning pipeline. Her AI Engineer World’s Fair 2025 keynote, “Spark to System: Building the Open Agentic Web,” brought this platform agenda to the conference.[3][1]

In October 2025, Sharma and Rolf Harms developed a more explicit argument about AI-product differentiation in “The Signals Loop.” Their proposed loop captures user interactions and product usage, then feeds that information into model refinement and product changes. They argued that fine-tuning was strategically important and that frequent model updates required changes to data pipelines, evaluation systems, and team workflows.[4]

The authors illustrated the loop with work by the Dragon Copilot and GitHub Copilot teams. Dragon Copilot used clinical data for fine-tuning, customer-feedback telemetry for refinement, and automated evaluation when new foundation models appeared. For GitHub Copilot completions, they reported over 30% improvement in retained code and 35% improvement in speed following both model-training changes and client-side and interface changes. The example combined model and product work; it did not isolate fine-tuning as the sole cause of the gains.[4]

Before her Microsoft AI-platform work, Sharma was Chief Operating Officer at Instacart and a Vice President at Meta. Her EmTech speaker biography describes broad Instacart responsibilities spanning product, design, data science, research, operations, and the P&L. At Meta, it describes her leadership of product groups working on Messenger, Instagram Direct, Messenger Kids, calling and video, and company-wide platforms. These roles supplied consumer-product and operating experience before her AI-platform and gaming appointments. The biography also records a Bachelor of Science in business from the University of Minnesota’s Carlson School of Management.[2][5]