AI Foundry
Accelerating foundational research into product innovation and driving customer value.
Self-Evolving Models & Agents
As agentic capabilities increase, AI systems will become self-evolving on both the model and harness levels. Our team is researching how to bring these self evolving capabilities into trusted enterprise use cases. This includes our self-improvement framework for harness evolution, as well as our SFR-AutoR&D and AutoInfra for Improving codebases, training models and advancing scientific discovery in AI Research.
Operational Intelligence
Operational intelligence works invisibly in the background, and turns business processes into reliable, agentic workflows. A great example is our foundational time-series model Moirai, which allows businesses to more accurately predict anything from contact center capacity to infrastructure demand. Read more about our vision for Operational Intelligence at Salesforce.
Ambient Intelligence
Ambient intelligence brings AI into the flow of work without interruption — context-aware systems that perceive, reason, and act at exactly the right moment. We’re conducting experiments, such as PISA which puts this into practice for live virtual calls and delivers relevant insights in real time. Read more about our vision for Ambient Intelligence.
Multi-Agent Systems
As we move from single agents to multi-agent systems, there are many important aspects to consider. From agents that coordinate and collaborate within an organization to agents that negotiate across business lines. Read more about our vision for Multi-Agent Ecosystems.
Simulation Environments
Reliable enterprise AI requires more than good models — it requires rigorous testing against the full range of conditions agents will face in production. Our team has launched eVerse, a simulation environment that exposes agents to thousands of edge cases, handoffs, and judgment calls before they ever reach a real workflow. Read more about our vision for Simulation Environments.