Fireside Chat
The Pilot Trap: Why Enterprise AI Stalls and How to Break Through | RAISE Summit 2026
- Industry adoption is projected to shift from pilot-heavy landscapes to production deployment, though many current agents remain effectively in pilots due to a lack of utility on the P&L line.
- First-generation agents are expected to continue deflecting tickets and saving minutes, whereas high-value agents in processes like underwriting or supply chain require cross-functional re-engineering to move significant financial value.
- Advanced use cases enabling human-inaccessible tasks, such as complex physical simulations and molecular discovery, are anticipated to deliver substantial value within two to five years.
- Jevons' paradox suggests that as intelligence costs decline, usage will expand to create more jobs in fields like personalized medicine and entertainment, with data indicating that AI spending correlates with faster headcount growth.
- Short-term automation is predicted to target coordination roles across functional areas, while organizational change management regarding acceptance, training, and complex enterprise processes is expected to remain the most difficult adoption hurdle.
- Governance is forecast to evolve from passive monitoring to real-time intervention and stopping mechanisms, starting long before agents run to ensure compliance and enable high-speed adoption similar to automotive safety features.
- Operational expectations for agents include establishing identity, accountability, and enterprise context training comparable to human employees, alongside capabilities for auditing, explanation, and performance-based "firing."
- Regulated organizations face the risk of reverting to low-value agents if risk, legal, and regulatory sign-offs slow down high-speed adoption, while a higher reliability bar is expected for agents than human counterparts.
- Infrastructure challenges related to trust and data are identified as primary barriers; specifically, the exhaustion of publicly generated data is viewed as the biggest impediment to advanced AI, despite open-weight models reaching parity with closed models in approximately six months.
- Future innovation is directed toward applying reasoning engines to domain-specific physical and chemical domains, including robotics and drug discovery, with continued investment in physical AI.