<p>A small cohort of engineers — Andrej Karpathy, Mitchell Hashimoto, Simon Willison — are producing in a week what used to take a month. Meanwhile, seventy-eight percent of enterprise AI deployments show no bottom-line impact. Ninety-five percent of pilots fail within six months. The gap between the people getting extraordinary results and the organizations getting nothing is not talent. It&#39;s architecture. And it has a name. </p><p><br></p><p>In this episode of the Product Impact Podcast, Arpy and Brittany break down why enterprise AI is failing at scale, what the engineers who are eighteen months ahead have figured out, and the two radically different futures that orchestration makes possible. </p><p><br></p><p>In this episode we cover:</p><ul><li>The $490 billion AI value crisis — why adoption is surging and returns are near zero, and what Forrester, McKinsey, PwC, and Gartner are documenting</li><li>Five failure patterns hiding inside every enterprise deployment — and why more training, more change management, and more executive support won&#39;t fix any of them</li><li>The pioneers building the future of work in public — Karpathy&#39;s vibe coding, Hashimoto&#39;s production-code throughput, Willison&#39;s hundreds of public experiments — and what they&#39;ve proven about orchestration as engineering discipline</li><li>Two outcomes of orchestration that most organizations aren&#39;t ready for: building bespoke deterministic software at a scale that was never economic before, and building an operating system where agents work autonomously on your behalf</li><li>Why markdown — not PDFs, not databases, not dashboards — is emerging as the knowledge substrate for the agent era, and why Karpathy himself is now calling for AI to organize wikis rather than chat</li></ul><p><br></p><p>&quot;These are not technology failures. They are failures of imagination about what work actually is and how AI fits into the way we work.&quot; — Arpy Dragffy</p><p>&quot;The primary failure mode in AI adoption is not capability. It is transferability.&quot; — Brittany Hobbs (citing Harvard Business Review)</p><p><br></p><p><strong>https://productimpactpod.com</strong></p><p>Thank you for listening to the Product Impact Podcast (formerly Design of AI) — Prove impact. Improve impact. Scale impact.</p><p>Hosted by:</p><ul><li>Arpy Dragffy Guerrero — <a href="https://www.linkedin.com/in/adragffy/" target="_blank" rel="ugc noopener noreferrer">https://www.linkedin.com/in/adragffy/</a></li><li>Brittany Hobbs — <a href="https://www.linkedin.com/in/brittanyhobbs/" target="_blank" rel="ugc noopener noreferrer">https://www.linkedin.com/in/brittanyhobbs/</a></li></ul><p><br></p><p>Go to Substack to get AI strategy frameworks, news, and jobs: <a href="https://productimpactpod.substack.com/" target="_blank" rel="ugc noopener noreferrer">https://productimpactpod.substack.com</a></p><p><br></p><p><br></p><p>This episode was brought to you by:<br><strong>PH1 (</strong><a href="https://ph1.ca/" target="_blank" rel="ugc noopener noreferrer"><strong>https://ph1.ca</strong></a><strong>)</strong> — an AI strategy consultancy specialized in improving the measurable success of AI products</p><p>.<br><strong>AI Value Acceleration</strong> (<a href="https://aivalueacceleration.com/" target="_blank" rel="ugc noopener noreferrer">https://aivalueacceleration.com</a>) — The consultancy specialising in enterprise value creation. Make sure that your spending doesn&#39;t go to waste. Find out exactly where the value creation of adopting AI products stalls.</p>