Experts: Reliable AI Requires Fresh Data and Clear Access Rules
Reliable, scalable AI applications depend on up-to-date information and clear access rules, according to speakers at Inc42's CTO Summit 2026. The challenge involves making data accessible from various sources while maintaining quality and control.
The foundation for reliable, scalable Artificial Intelligence applications relies heavily on access to fresh data and clearly defined usage rules, industry leaders stated at Inc42’s ‘The CTO Summit 2026’ in Bengaluru. The core challenge discussed was making information accessible across disparate databases, data warehouses, and documents without compromising data quality or control over its use.
Panellists, including Kisalay Ranjan, head of engineering at PhonePe; Bhanu Jamwal, GM of India business at OceanBase; and Anil Sinha, CTO of Fibe, explored strategies for managing this data flow. Ranjan highlighted PhonePe’s approach of tracking data origins and setting granular usage rules, emphasizing that retrofitting these controls at the application level becomes increasingly difficult with higher automation.
Jamwal pointed out the common issue of data being spread across separate systems for transactions, analytics, and AI. He advocated for a unified data layer to enable AI agents to quickly retrieve current information. Fibe, a lending tech startup, shared its use of generative AI to extract signals from diverse documents to enrich its credit model, rather than for direct lending decisions.
Beyond data quality and accessibility, the summit also addressed the trade-offs between AI model performance, cost, and response times. Speakers suggested that the choice between smaller, faster models for high-volume tasks and more advanced models for complex reasoning should be dictated by the specific use case, model availability, and economic factors.