Automation AppThe Automation Company is your single source for RevOps and Marketing Automation solutions. We empower businesses to achieve more, faster.

Why Salesforce is shifting away from model-based AI

26 May 2026 · 1 min

Watch on YouTube

0:32

In this brief presentation, two Salesforce speakers explain the company’s strategic move from focusing on the "model layer"—the individual AI models—to a broader "system‑based AI" approach. They argue that models alone cannot deliver best‑in‑class agents for customers; instead, AI must be built as an integrated system of components, much like a car’s engine and its surrounding parts. This system‑based view lets Salesforce address human‑centric, business‑problem‑driven use cases, while still being grounded in the foundational research that underpins the technology. The shift emphasizes infrastructure, component ensembles, and a focus on solving real‑world enterprise challenges rather than merely improving isolated models.

For the past few years, we've been really focused on what we call the model layer.
Uh, models themselves are not sufficient for building these best-in-class agents, for building the platform that is critical for our customers.
And now, uh, we're able to actually move, uh, one abstraction layer above that.
So today we're shifting from, uh, model-based AI into system-based, uh, AI.
What that means is that we can think about the business problems, about the human-centric, uh, problems that AI can solve.
So when you build a car, you not just, you know, use the engine, you have to actually build all the components which are connected to the engine.
So in a way, by system-based AI, we look at the ensemble of components which are critical, but it needs to be really driven or fed by the foundational research that the team is working on, uh, and researching every day.
Salesforce, we are really building, uh, the whole infrastructure and all the system around those models such that this, this agents can be really best in class and, uh, extremely, uh, consistency in their performances. We're not just scientists in lab, but we're also building things, and we're very excited to see how these new things we're building will transform the business and the enterprise AI.