From hallucinations to hard truths
summary
A field report on what actually breaks and what fixes it: demos that went nowhere after rushing to build; the first AI feature that takes six months to launch and the pivot to embed AI engineers into product teams that finally ships; governance rules like PII never touches the model and an eval framework tying model accuracy to cost, revenue and latency; Gartner warnings—85% sponsor failure and ~40% agent projects cancelled—and a six‑figure cost variance from unoptimized models.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 0:50Solving the wrong problem
- 2:14Impact versus effort
- 2:35Lack of executive sponsors
- 3:59Team not built to ship
- 5:25Hype around agentic AI
- 6:51Ignoring AI evals for ROI
- 7:54Biggest risk is not the model
- 9:18Model slow and expensive
- 10:21Summary
related talks
37mPanel Discussion: Developing with AI
52mAI Is Just Software, What Could Possibly Go Wrong?
58mThe Better Quality Movement - How We Can Advocate for a Better Digital World
36mAvoiding 5 missteps that undermine your AI readiness and success
39mFrom Demo to Production: Why Agentic AI Systems Fail - and How to Fix Them
10mNavigating the AI Code Explosion
56mPANEL: Unleashing the Future: The Impact of Gen AI on Development Productivity
14mAI Sprawl: The Hidden Risk Threatening Your Enterprise Now