Salesforce's Benioff on Enterprise AI: Slow Adoption, But Promising Future

Salesforce Chairman and CEO Marc Benioff spoke at The Six Five Summit: AI Unleashed 2026, arguing that enterprise artificial intelligence is moving from hype to execution, albeit slowly. He said many large companies are only just beginning their AI work while continuing to spend on old systems.

Benioff described a three-phase approach to AI adoption in the enterprise: prompt-based models (AI 1.0), agentic systems that can carry out tasks independently (AI 2.0), and dynamic interfaces that merge deterministic enterprise software with probabilistic AI tools (AI 3.0). He said the third phase is where real value begins to show up, allowing users to interact with data and applications in a more flexible way.

Salesforce is already building toward this model through its 'fully headless system', with applications built in Lightning and rendered through new APIs. Benioff highlighted several AI systems currently running inside the company or in customer environments, including Casey (5 million customer service issues resolved) and Slackbot (supporting tens of millions of Slack users).

Benioff emphasized that companies should not think of AI as a reason to discard existing systems, but rather to unlock value already trapped inside them. He called this the 'happy middle', where AI is added on top of deterministic core systems to provide control and flexibility.