AI Gets You 60% of the Way. The Last 40% Is Where Everyone Fails | KPMG X LYZR
Anyone can build an agent demo now. Almost nobody can scale one. In this episode of How We Agent, Ani sits down with Swami Chandrasekaran, Partner and AI leader at KPMG, who has been through four generations of AI and now spends his days figuring out how a 125-year-old firm of tax, audit, and advisory professionals actually reinvents its work. Swami is refreshingly blunt about the industry's vocabulary treadmill, from frameworks to harnesses to whatever comes next, and equally blunt about what he looks for before he takes a startup call. The conversation covers building an agent harness from scratch to understand it, productizing professional services, the criteria KPMG used to evaluate agent platforms, what actually goes wrong at scale, and why trust is the thing that drives adoption. In this episode you'll learn: Why Swami builds things from scratch instead of reading about them Frameworks, harnesses, and why the industry renames the same thing every 18 months How KPMG's AI mandate shifted from experimentation in 2023 to scaling in 2026 What it means to productize a professional services business How KPMG built on top of raw model intelligence rather than buying off the shelf The criteria KPMG used when evaluating agentic platforms 24 months ago Three things that get a startup on an enterprise buyer's radar His personal smoke test: are you brave enough to let people taste the ice cream Why the last 40% is security, trust, scale, and actual adoption How AI goes wrong when you automate bad processes instead of reimagining them Why mid-market and enterprise are both suddenly open to everyone What Swami is teaching his kids about curiosity and creative thinking ⏱ Chapters [00:00] The Last 40% Is Where It Gets Hard [00:25] How We Agent: Building Agents Inside KPMG [00:42] Why Swami Built an Agent Harness From Scratch [02:06] Frameworks, Harnesses, and the Vocabulary Treadmill [03:22] Four Generations of AI and a Blank Sheet of Paper [04:15] The Three-Part Mandate: Innovate, Enable, Standardize [05:36] Building an Ecosystem Instead of Building Everything [06:13] Why KPMG Is Now in the Scaling Phase [06:34] What KPMG Built Internally vs Adopted Outside [07:03] Is Productization a Pivot or a New Revenue Channel? [08:43] Digital Gateway, CLA, and Agentizing the Core Platforms [10:17] How KPMG Decides to Work With a Startup [11:32] Are You a Journey Partner or a Me-Too Product? [12:18] How to Get on an Enterprise Buyer's Radar [13:33] The Smoke Test: Taste the Ice Cream [14:49] What's the Right Thing to Solve Right Now [15:16] Scale, Security, Trust, and the Last 40% [16:28] What Could Go Really Wrong With AI [17:44] RPA 3.0: Automating Bad Work With Better Tools [18:29] Trust Builds Adoption [18:55] What Could Go Right: New Markets, New Professions [20:12] Advice for the Next Generation: Be Curious, Be Creative [21:30] Curiosity Only Counts If You Build Something With It Enterprise AI, Agentic AI, KPMG, AI Harness, Agent Frameworks, Professional Services, AI Scaling, Enterprise AI Adoption, AI Trust and Safety, AI Guardrails, Productization, Startup Enterprise Sales, Multi-Agent Systems, Digital Transformation, Future of Work 🏷 YouTube Tags Enterprise AI, KPMG, Agentic AI, AI Agents, Swami Chandrasekaran, Lyzr, How We Agent, AI Harness, Agent Frameworks, Enterprise AI Adoption, AI Scaling, Professional Services AI, AI Trust, Startup Enterprise Sales, Artificial Intelligence 🔗 Important Links: Build with Architect: https://hubs.ly/Q043pWTs0 Build your own AI agent → https://hubs.ly/Q03wb5Md0 Explore our website → https://hubs.ly/Q03wbGVt0 Build agents for your company (Book a demo) → https://hubs.ly/Q03wbH0k0 Learn how to build agents with Lyzr Academy → https://hubs.ly/Q03wqxFR0