As generative AI enters the end of its honeymoon phase, organisations are moving beyond experimentation and preparing to commit. However, turning early AI use into widespread deployment remains a significant hurdle for many startups. The biggest challenges are operational — increasing costs, governance and trust issues, and the transition from individual AI use to managing autonomous digital workers.

These were the key talking points during a recent Sifted Talks session in partnership with intelligent content management platform Box, where panellists unpacked how founders can break out of the pilot phase and build resilient, AI-native operations. The discussion focused on practical strategies for scaling AI responsibly while addressing the operational complexities that emerge when moving from proof-of-concept to production-grade systems.