AMA: Salesforce Director of Product Management, Tableau Next, Anushka Anand on AI Product Management

We use thumbs up/down signals as a lightweight always-on feedback layer, but the real signal comes from betas and pilots with targeted customers before broad rollout. That direct feedback loop early is how we tune AI features before they're in everyone's hands. The most consistent theme we hear: users want the AI to know their world — their data, their process, their requirements. Generic intelligence isn't enough. The feedback keeps pushing us toward more customization and control. The more an ...Read More

AI has genuinely shifted where I spend mental energy — less on mechanical work, more on judgment-heavy stuff. For research, I can get from "I know nothing about this space" to "informed enough to have real conversations" way faster. For brainstorming, I use it as a thinking partner — doing quick market research passes, understanding what customers love or hate about existing solutions, pressure-testing ideas against our differentiators before bringing anything to the team. And the tools make it ...Read More

Honestly, the "could be AI" trap is real. Our filter is pretty simple: does this move us closer to making customers wildly successful with agentic analytics, or is it just cool? We're working on a genuinely hard problem — building natural language experiences, deeply integrated in their work of work, that reason about data and produce insights people actually trust. The hallucination and non-determinism challenges mean that "good enough" AI isn't good enough for us. That bar helps us say no to a ...Read More

The "ask me anything" type of agent is becoming a baseline expectation across various workflows. But the use cases actually getting traction aren't standalone chatbots — they're contextual, in the flow of work. At Salesloft, sellers were asking agents everything from call prep guidance to "what does my pipeline data say about where I should focus." In my current role, business users want to ask follow-up questions right from the dashboard they're already looking at — no context switching, no pin ...Read More

Trust is the unlock. The AI workflows I've seen get real adoption all have one thing in common — users understood what the agent was doing and why, before they were asked to rely on it. Concretely that means showing your work. Letting users see how the agent reasoned, what plan it's executing, where it's confident versus uncertain. And giving people a way to calibrate and test before they go fully autonomous — so trust is earned incrementally, not assumed. The other big pattern is that workflows ...Read More