Salesforce: building its AI product roadmap through weekly customer sessions

On April 30, 2026, TechCrunch reported that Salesforce is developing its AI product roadmap through direct, ongoing sessions with enterprise customers — in some cases meeting weekly — rather than relying on standard quarterly planning cycles.

The company works across 18,000 customers organized around shared problem themes, such as “agent context” and “observability.” The organizing principle is that if one enterprise customer faces a problem, others likely encounter the same one. Solutions validated with a specific customer are then evaluated for broader release across the customer base.

Two examples from the article illustrate how this works in practice. Travel platform Engine meets with Salesforce weekly and provides feedback that directly influenced refinements to Salesforce’s AI voice agent. Federal credit union PenFed developed an IT service management workflow using Agentforce that Salesforce subsequently packaged and released more broadly.

The approach trades the predictability of structured roadmapping for speed and signal quality. A company meeting weekly with customers gets validated insight about what works and what does not far faster than one operating on quarterly planning cadences.

The article also identifies a structural risk: relying heavily on enterprise customer judgment to set product direction may not produce decisions that hold up as market conditions shift, since customers are better at describing problems they face today than anticipating problems they will face in twelve months.

Why it matters for product managers

This is a live example of continuous discovery applied at enterprise scale to an AI product. For product managers working on AI features inside enterprise software, it illustrates how tight customer feedback loops can substitute for speculative roadmapping in a domain where both the technology and the use cases are changing quickly. The model works because Salesforce can treat individual customer problems as signals about the broader market — a condition that holds when the customer base is large and varied enough that shared problems emerge reliably.