What does it really take to move from AI hype to something that actually works inside a business?

In this episode, I sit down with Shibani Ahuja, SVP of Enterprise IT Strategy at Salesforce, to talk about why so many enterprise AI projects stall long before they deliver real value. While the market is full of noise around agents, copilots, and automation, Shibani makes the case that the real issue is often much simpler and much harder at the same time. Design. She explains why model capability alone will never rescue poor architecture, weak governance, or unclear data ownership.

Our conversation goes well beyond the usual agentic AI headlines. Shibani shares what she learned from speaking with hundreds of C-suite leaders over the past year, and why many early enterprise AI conversations were too focused on models instead of ecosystems. We unpack the difference between predictive, generative, and agentic AI, why trusted data means more than having lots of information, and how Salesforce's own internal journey revealed conflicting knowledge, governance gaps, and the importance of determinism in enterprise settings.

I also loved Shibani's perspective on the human side of this transformation. We talk about why successful organizations are framing agents as a capacity multiplier rather than a headcount story, how to bring employees along through visible wins and shared learning, and why the best starting point is often a simple, boring use case that removes pain for frontline teams. She also shares her thoughts on the eight design principles for the agentic enterprise, the myths that frustrate her most, and what will separate the leaders from the laggards over the next 18 to 24 months.

This is a conversation for anyone feeling pressure to do something with AI, but wanting a clearer view of what meaningful progress actually looks like. Are businesses building the right foundations for an agentic future, or are too many still mistaking experimentation for strategy? Have a listen and let me know your thoughts.