Salesforce is letting customers drive its AI roadmap — and that’s smarter than it sounds

Salesforce has always had a knack for turning customer feedback into product features. But with AI, they’re taking it a step further: they’re basically letting their customers decide what gets built next.

The logic is refreshingly straightforward. If one big enterprise customer hits a wall with Salesforce’s AI tools — say, the agentic features aren’t handling a specific workflow, or the data integration is clunky — chances are good that other customers are bumping into the same problem. So instead of relying solely on internal product managers and data scientists to guess what’s needed, Salesforce is actively crowdsourcing the roadmap.

This isn’t just a feel-good “we listen to our customers” PR move. It’s a pragmatic response to the chaos that is enterprise AI right now. Every company is experimenting, but nobody has a clear playbook. Salesforce’s customers range from banks to retailers to healthcare providers, and each one has a different set of data governance rules, compliance requirements, and actual business workflows. No internal team can anticipate all of that.

By letting the customers lead, Salesforce essentially turns its user base into a massive, distributed R&D department. When a customer proposes a feature, Salesforce can validate it against other similar customers, prioritize it, and ship it faster than if they were guessing. The downside? You might end up with a roadmap that’s a bit reactive, and maybe not as visionary as something cooked up by a team of AI researchers. But honestly, in the current market, “reactive to real problems” beats “visionary but irrelevant” every time.

I’ve seen this approach tried before in other SaaS companies, but usually at a smaller scale. Salesforce has the advantage of a huge, diverse customer base and the engineering muscle to actually deliver. The risk is that they’ll over-index on the loudest voices — the biggest customers with the most leverage — and ignore the needs of smaller players. But if they can keep the feedback loop balanced, this could be one of the smarter AI strategies out there.

The real test will be whether the features that come out of this actually feel cohesive, or if the product ends up as a Frankenstein’s monster of customer requests. I’m cautiously optimistic. Salesforce has been around long enough to know that you can’t just build everything everyone asks for. But letting the customers steer the ship, at least for now, seems like a solid bet.

And honestly, I’d rather have a product roadmap shaped by people who are actually trying to use the thing in the real world, day in and day out, than by a bunch of product managers in a San Francisco boardroom. That’s not a knock on product managers — it’s just a recognition that enterprise AI is still so new that nobody has all the answers. The customers might not either, but they definitely know which questions hurt the most.