Salesforce’s AI Journey: Evolution or Constant Reset?

Over the past couple of years, I’ve been closely following Salesforce’s AI journey. And honestly, I’m struggling to answer a simple question: is this a continuous evolution, or a constant reset?

A New Vision, Every Time

Since the rise of generative AI, almost every Salesforce keynote has introduced a new vision, a new layer, or new terminology.

Each one sounds significant. Each one sounds like the future. But they do not always feel as though they are clearly built on top of one another.

The AI Narrative Keeps Changing

The change is not limited to product names; the underlying AI narrative has shifted as well. At one point, the focus was OpenAI. Then Gemini Flash was presented as the future. At TDX, Claude took the spotlight.

The pattern is familiar: a new model arrives, followed by a new narrative.

What About the Core Vision?

This is where the more important question begins. Enterprise organizations do not simply buy features, demos, or keynote excitement. They invest in clarity, consistency, and a direction they can trust over time.

Watching the latest TDX keynote, one point stood out. Salesforce is clearly moving toward agents, control layers, and hybrid systems that combine deterministic and probabilistic approaches. That direction makes sense.

At the same time, it raises a fair question: wasn’t the earlier narrative about removing complexity—replacing workflows, eliminating rules, and allowing AI to handle everything?

Now We Are Back to Structure

Today, the message is different. Organizations are being asked to define conditions, control flows, test outcomes, and manage behavior. In short, AI needs rules.

It also reveals something important: the industry did not fully know where this journey was heading. What we are seeing is not a perfectly linear roadmap, but real-time learning at scale. Even the largest vendors are experimenting, repositioning, rephrasing, and recalibrating.

The Real Enterprise AI Challenge

I do not think Salesforce is lost. But I do think the AI story is still being written—even by the biggest players. And perhaps that is understandable.

The real challenge is not building impressive demos. It is building systems that work reliably in real business environments.