Robust reasoning, multi-modality, and collaborative agent systems are the key advancements shaping the future of AI agents

On the sidelines of the TrailblazerDX 2025 event, San Fransisco, Shipra Sinha, Lead Analyst - Industry Intelligence Group (IIG), CyberMedia Research (CMR) sat down with Jayesh Govindarajan, EVP Software Engineering at Salesforce. Jayesh leads the Salesforce AI org and his team is tasked with setting strategy, designing, building and operating Predictive, Generative and Agentic AI systems.

- By Shipra Sinha

How do you see Predictive, Generative and Agentic AI working together to transform business decisions?

Salesforce has been on the predictive journey for eight or nine years now and doing about a billion predictions a day and the way that impacts the enterprises is infused in almost every single component that's on Salesforce. Looking at predictive AI some functions and components are probabilistic and help with predictions that are deeply tied to a flow, a persona, in sales, marketing, service and in almost everything. Generative AI was launched two years ago with prompt builder and prompt engines which enable customers to bring generative AI into the flow of work. Salesforce has about 2500 customers, doing about 83 million generations a month on the Salesforce platform. The new wave is the agentic wave, which is more conversational, more multiturn, has multi-step reasoning built in, access to all the actions in the ecosystem, access to data, and cloud and brings all of that together into an agent. Agentic AI is far more advanced, has far more agency in taking actions on your behalf, can orchestrate, bring in the right data, and orchestrate the right actions. Anybody can build an agentic system, wrap it into an API and be able to use it anywhere with the front end which is conversational or connect it to the back ends that they want to automate.

What is the biggest challenge of integrating AI into the existing enterprise systems and how did you address them?