Salesforce Hands the Wheel to Its Customers
In an era where tech giants typically build in secret and ship to fanfare, Salesforce is trying something different: asking the people who actually use its software what it should build next.
The company has formalized a customer-led approach to its AI product roadmap, creating a feedback loop where enterprise clients surface real-world pain points — and those problems get prioritized directly on the engineering backlog. The core logic is simple but compelling: if one large enterprise customer is struggling with something, the odds are good that thousands of others are facing the same friction.
Why This Matters for Enterprise AI
Most software companies collect customer feedback through support tickets, NPS surveys, and the occasional advisory council. What Salesforce is describing goes a step further — positioning customers not just as feedback-givers but as active co-authors of the product direction.
This approach makes particular sense in the AI space, where the gap between what a tool can do and what it actually does reliably in production is still enormous. Enterprise AI deployments are notoriously messy. Data is siloed, workflows are idiosyncratic, compliance requirements vary by industry, and hallucinations can cause real business damage. No amount of internal testing fully replicates that chaos.
By tapping into live enterprise deployments for signal, Salesforce gets something valuable: a continuous stream of failure modes and edge cases that would take years to discover in a lab.
Agentforce at the Centre
This customer-crowdsourcing strategy is tightly linked to Salesforce's push around Agentforce — its platform for building autonomous AI agents that can handle tasks across sales, service, and operations without constant human hand-holding.
Agentforce has been Salesforce's most prominent AI bet heading into 2026. But autonomous agents interacting with real business data, real customers, and real workflows create a whole new category of failure scenarios. Getting direct customer input on where those agents break down — and what they'd need to actually be trusted — gives Salesforce a roadmap that reflects the market's real readiness, not just its ambitions.
A Smarter Way to Build?
There's a competitive dimension here too. Microsoft, Google, and a wave of AI-native startups are all competing for enterprise AI budgets. Salesforce's existing relationships with hundreds of thousands of businesses are a moat — but only if the product keeps earning trust.
Crowdsourcing the roadmap is one way to keep the product anchored to actual customer problems rather than drifting toward whatever looks impressive in a demo. It also creates stickiness: customers who feel heard tend to stay.
Whether this approach produces meaningfully better AI products than the traditional closed-door R&D model remains to be seen. But in a space where enterprise buyers are still deeply skeptical of AI's reliability, showing that your roadmap is shaped by real-world deployments rather than boardroom vision is itself a credibility signal.
For now, Salesforce is making a calculated bet: the best people to design AI for enterprises are the enterprises themselves.
Source: TechCrunch — Salesforce is crowdsourcing its AI roadmap — with customers
