Salesforce Chief Executive Officer Marc Benioff outlined a three-phase framework for enterprise artificial intelligence adoption during a keynote at the Six Five Summit’s AI Unleashed 2026 event on August 25, positioning the company to transition from prompt-based models to agentic systems and dynamic interfaces within the next 12 to 18 months.
Benioff described the first phase, AI 1.0, as the public data-driven prompt era exemplified by tools such as ChatGPT, which he characterized as occurring roughly three to four years prior to the 2026 summit. The second phase, AI 2.0, involves autonomous agents capable of executing tasks, including Salesforce’s Agentforce platform, customer service agent Casey, Slackbot, and supply chain agent Marshall. The third phase, AI 3.0, envisions dynamic interfaces that integrate deterministic enterprise software with probabilistic AI models, with Benioff stating the company aims to deliver such an interface within 12 months.
The company reported nearly $43 billion in revenue over the last twelve months and a market capitalization of $169 billion, with a gross profit margin of 77.6%, revenue growth of 11%, a PEG ratio of 0.61, and a free cash flow yield of 9%. Salesforce’s stock closed at $209.17 on August 25, up 1.82%.
Benioff emphasized the gap between enterprise AI hype and reality, noting that many customers remain in early stages of transformation despite widespread discussion. He cited personal meetings with over 200 customers during eight weeks in Europe, highlighting concerns about integrating AI with existing enterprise systems such as Salesforce, Microsoft, SAP, and proprietary data lakes. The company has hired thousands of new graduates to assist in building governed AI systems, aiming to provide controlled, secure development environments for enterprise use.
The executive also referenced Dell’s Regrello system, which interacts with 20,000 suppliers daily for status updates and compliance checks, as an example of agentic AI in operational workflows. Benioff cautioned against relying solely on vendor narratives, stating that customers should evaluate systems independently rather than accepting market claims at face value.
