Salesforce is letting customers steer its AI roadmap — and honestly, that makes sense

Salesforce has been around long enough to know that the loudest voices in enterprise software aren’t always analysts or VCs — they’re the customers who actually use the thing day in, day out. So it’s no surprise that the company is now leaning hard into crowdsourcing its AI roadmap directly from those customers.

The logic is simple: if one large enterprise customer is banging their head against a specific problem, chances are a dozen others are too. Instead of guessing which AI features to build next, Salesforce is effectively saying, “Tell us what you need, and we’ll prioritize it.”

This isn’t entirely new. Salesforce has always had customer advisory boards and user groups. But the shift here is the scale and the focus on AI. The company is actively soliciting input on which use cases to tackle — things like automating complex workflows, improving predictive lead scoring, or making natural language queries actually work inside CRM without hallucinating half the response.

What I find interesting is the timing. We’re past the peak of AI hype, where every vendor was throwing generic chatbots at the wall to see what stuck. Now we’re in the phase where companies have to prove ROI, and nothing proves ROI faster than solving a real problem a paying customer already has.

Of course, there’s a catch. Crowdsourcing works great when your customer base is large and diverse enough to surface common pain points. But it also risks turning the roadmap into a popularity contest. The biggest customers — the ones spending millions — will naturally have more sway. Smaller businesses might find their niche issues deprioritized unless they’re loud enough or lucky enough to align with a trend.

Salesforce has been here before with its IdeaExchange platform, which has a mixed track record. Some ideas get traction and ship quickly; others languish for years. The difference this time is that AI moves fast, and customers expect faster feedback loops. If Salesforce can actually close the loop — show customers that their input led to a shipped feature — this could build serious loyalty. If not, it’s just another suggestion box.

I also wonder about the internal tension this creates. Product managers usually have a vision, a roadmap, and a sense of where the market is heading. Handing the wheel to customers can lead to incremental improvements rather than bold leaps. But maybe that’s exactly what enterprise AI needs right now: less moonshots, more boring stuff that actually works.

One more thing: this approach puts pressure on Salesforce’s AI models to be flexible. If customers can request custom workflows or domain-specific behaviors, the underlying AI has to support that without breaking everything. That’s a technical challenge that’s easy to underestimate.

All told, I’m cautiously optimistic. Letting customers lead on AI feels like a mature move from a company that’s been through enough hype cycles to know better. The real test will be in execution — whether they can listen, prioritize, and ship faster than the competition.