David Bombal #565: Stop the AI Hype: What Enterprise Teams Are Really Building
Mar 25, 2026
Carlos Pereira, Cisco Fellow and Chief Architect for Customer Experience, explains why enterprise value comes from agentic AI that executes B2B workflows, not consumer chatbots. He walks through how teams narrow hundreds of ideas to a few ROI-backed use cases. Discussion covers evaluations, input guards, latency and data sovereignty, and how embedding AI into workflows drives real adoption.
AI Snips
Chapters
Transcript
Episode notes
Define Funding Criteria Before Choosing AI Use Cases
- Do define strict funding criteria before picking AI use cases so investments target measurable ROI.
- Carlos built three criteria (maximize Cisco tech value, improve customer resilience/security, enable proactive personalized interactions) and cut 412 use cases to 5 with proven ROI.
From 412 Ideas To Five Funded Use Cases
- Anecdote: a European customer handed Cisco a spreadsheet with 412 potential AI use cases.
- After Cisco's advisory work they prioritized five with real ROI and two IT implementation paths for scale.
2025 Was The Year Of Evals Not Just Hype
- Evaluations became central in 2025 as teams learned to measure probabilistic models against deterministic needs.
- Carlos calls 2025 both the year of 'gen take' and the 'year of evals' when hallucinations forced performance testing.