Many AI agent projects stall after promising pilots because they can’t deliver the accuracy businesses need to trust them. A key reason is that enterprise data isn’t just numbers and tables- it comes with strict privacy and governance rules as well as team-specific terms, definitions, and processes. When agents are given access to this data without a system that can enforce governance policies or adapt to evolving definitions in real time, they quickly run into the “accuracy trap”: results that are confidently wrong, non-compliant, or misaligned with how the business actually works. This talk shares lessons from early enterprise deployments on why accuracy debt and weak governance hold back adoption- and how leading teams are building AI agents that respect privacy, adapt to governance, and continuously learn from tribal knowledge to deliver reliable, explainable results.
Rajoshi Ghosh
Rajoshi Ghosh is the Co-founder and Chief Ecosystem Officer at PromptQL, where she works across product, marketing, and sales to bridge the gap between research and real-world impact in AI. She brings PromptQL to AI change agents driving transformative projects—helping them build reliable, custom AI for their most critical data workflows.
Rajoshi has deep 0–1 startup experience in bringing disruptive technologies to new markets, including the launch of the Hasura GraphQL Engine. Before Hasura, she ran a technology consulting firm helping companies modernize infrastructure with containerized workloads. Earlier in her career, she was a bioinformatics researcher, with her work published in Nature.