Tech Unplugged

Chief of staff is a role a lot of people at LBS are curious about, and almost nobody can describe properly. So we asked someone who has done it twice. Ally O'Donohoe is Chief of Staff and Director of Ops and People at Fyxer AI, and was chief of staff at Ukio before that. Three things stood out.There is no single version of the job, but there is a shape to it. Ally described four pillars: leadership and exec programming, investor relations, strategy, and people. Which pillars you actually own depends on who is already in the building. If the company has a strong CFO, you do less on the investor side. If it is scaling fast, the people work expands to fill your week.Being a generalist used to be the trade-off. It is less of one now. You give up functional depth, and a lot of the work is behind the scenes with little external credit. But Ally made the point that getting up to speed on an unfamiliar area used to take months and now takes days, which makes breadth far more useful than it was a few years ago.How to spot a good role, and a bad one. If calendar and inbox management appear in the job spec, it is an EA role with a different title. And founder fit matters more than people expect, because you are in that person's pocket daily through the best and worst weeks. You are interviewing them as much as they are interviewing you.On getting in: cold outreach to founders, weekly, beats structured job hunting. And do not fixate on the title, since strategy and ops roles often ladder into it.

Anurag Viswanath has built product across Google, Deel and now Meta, spent a stint as a Founder-in-Residence at Antler, and started out as an officer cadet in the Indian Army.We sat down with him to talk about what has actually changed in product management now that AI is in everything, and what has just been repackaged.A few things that stood out:Forget T-shaped. Anurag went M-shaped. He deliberately stacked depth in two or three domains alongside breadth, and says the real payoff has been pattern detection: spotting the same structural problem across completely different surfaces."AI product manager" means at least four different jobs. From using AI to speed up your own work, to layering AI onto an existing product, to building AI-native from scratch, to the emerging agent PM role. He explains why picking the right flavour matters more than chasing the title.The 3B framework for breaking in. Borrow your domain expertise, Build something real, Benchmark its quality. He is honest that breaking in from a non-technical background is hard right now, and equally clear about why an MBA background is more of an advantage than people think.Taste and judgement cannot be shortcut. Evals can be taught. Judgement comes from reps. He shares a simple exercise for building it if you do not have years behind you.Agentic commerce is the tailwind he would go deep on. When an agent is doing the searching and choosing, there is no click left to win. He unpacks what that means for how you build, price and position a product.On leaving big tech for Antler: "In a big tech environment, a lot of your effectiveness is borrowed." He wanted to find out what happened when the scaffolding fell away.

Most of us follow AI by reading about it. We keep up with the news, skim the threads, and feel reasonably on top of things. Melina's argument is that this is exactly the trap: you can read endlessly and still have no real sense of where these tools actually work and where they quietly fall apart.For our latest Tech Unplugged episode, we spoke with Melina, a product leader who has worked across Meta, Coinbase, and Mistral, and now runs her own product advisory and teaches AI product management.A few things we got into:•⁠ ⁠Why the gap between a cool demo and something you actually use every day is where most of the real work (and value) lives•⁠ ⁠What "AI product management" actually means, and why most people only think about a third of it•⁠ ⁠How you build credibility with engineers when you can't explain how a transformer works, and why trying to can backfire•⁠ ⁠What starts breaking inside a company during hypergrowth, from knowledge silos to decisions grinding to a halt•⁠ ⁠Why a large code base no longer protects you, and what does: UX, data feedback loops, sheer execution speed, and operating in regulated spacesIf you're trying to get into AI, lead an AI roadmap, or just want a clearer picture of what's hype and what isn't, we think you'll get a lot out of this one.Have a listen, and let us know what you think.

Most companies running AI pilots aren't short on ideas. They're short on pilots that survive contact with a real workflow. In this episode of Tech Unplugged, I'm joined by Akshay Nagpal, who launched a machine learning platform at ING that became the second most adopted product in ING Analytics, and who now works in GenAI product and advises startups on conversational AI.We get into the parts people usually skip. What actually kills an AI initiative inside a big bank. Why COBOL still sets the ceiling on what's possible. And why the most valuable piece of automation Akshay shipped at ING had no AI in it at all. We also talk honestly about jobs — his view is that the "average analyst" role is already gone, and that nobody has built a real transition plan for the people in it.What we coverInnovation theatre: why pilots that don't start from a workflow or a revenue line quietly dieThe Biking initiative: automating year-end audit letters across Poland, the Netherlands, France and Italy, cutting weeks of manual coordination — with zero AI involvedThe 60-cent problem: legacy banks spend around 60 cents to earn a dollar. Monzo and Revolut spend 30–35. AI-native banks could reach 15–20Patriot Act vs GDPR: why European banks took years to trust AWS and GCP, and why Akshay defends GDPR as a consumerBuild vs buy: when partnering with OpenAI or Anthropic beats internal tooling, and the two cases where building your own is the right callWhen agents hire agents: two coding agents split backend and frontend, ran their own retrospective, and decided they needed a PMBeyond the wrapper: why the moat is now user experience, distribution and vertical embedding — not the technologyThe uncomfortable bit: which roles go first, and what "upskilling as an orchestrator" actually means in practiceTools mentioned in this episodeLovable to vibe-code expense tracker and portfolio siteClaude for redesigning onboarding flowsGamma for presentationsGemini for deep researchn8n and Zapier to run workflow automation (e.g. Stripe transactions straight into Slack)Chapters[0:00] Innovation theatre and the reality of scaling AI[2:15] From Nokia to ING to entrepreneurship[7:45] The Biking initiative: automating audits without AI[10:50] COBOL, technical debt and the 60-cent efficiency ratio[14:15] Build vs buy, and the ego of internal tooling[17:30] The Patriot Act, GDPR and why EU banks feared US cloud[20:45] Specialised agents: when AI hires its own PM[28:30] Moats, wrappers and vibe coding[32:15] Displacement, fresh graduates and the missing transition plan[39:00] Closing advice: have a bias for actionGuest: Akshay Nagpal — big data engineer at Nokia, then product manager at ING in Amsterdam, where he launched a machine learning platform that became the second most adopted product in ING Analytics with 22 B2B clients onboarded. Oxford MBA '24 and former Head of the Oxford Tech Club. He now works in GenAI product at Hark Labs and advises startups on GenAI workflows and conversational AI.Host: Aritra Sutradhar — LBS MBA '26. Consultant at BCG London. Previously chief of staff and GTM roles in startups, and years building in education. Speaks and writes on creativity in the age of AI.Source: Tech Unplugged — the London Business School Tech Club podcast

AI is no longer just a feature layer; it has become the product, the teammate, and, in some cases, the operator. But what does this shift mean for the people defining these products?In this episode of Tech Unplugged, we sit down with Amit Pasupathy, Senior AI Product Manager at Decagon, a company recently valued at $4.5 billion that is redefining customer experience through AI-powered concierge agents. Amit shares his transition from a Data Scientist at Freenome to an "Agent PM" at Scale AI and Decagon, exploring the technical and cultural nuances of building autonomous agents at scale.We dive into:Defining the AI Agent: Why "resolving" an issue is the critical differentiator between a true agent and a standard chatbot.The "Agent PM" Role: How product management is evolving into a "forward-deployed" role that requires building alongside engineers and customers.Multilingual Complexity: Navigating the "hair-whitening" challenges of international deployments, from GDPR compliance to the linguistic nuances between Norwegian and Danish.The Future of Work: Why your future AI agent will be your "colleague" rather than just a tool, and how role requirements are shifting from execution to auditing.Career Tactics: Why the "standard MBA path" to PM is changing and why building real-world, impactful projects is the new credible signal.Timestamps(00:00) Intro: AI as the Product(01:53) Amit’s Journey: From Stanford Math to Columbia MBA(03:52) Agentic AI vs. Chatbots: Defining the "Resolution" Layer(06:40) The Complexity of Multilingual Voice Agents(09:30) Case Study: Deploying Multi-Language Support for Fintech(11:58) Defining the "Agent PM": Forward-Deployed Leadership(15:15) The Scale AI Transition: Navigating Layoffs and Career Resilience(18:25) Is Technical Fluency Required? The Rise of "Vibe Coding"(21:05) Building with Impact: Advice for Aspiring PMs(23:10) Hiring at Decagon: The 3 P's (Product, Performance, Partnership(25:40) MBA Experience: CBS vs. LBS and the Big City Life"An AI Agent isn't just about providing information. It's about resolution. Can it actually resolve the customer's issue in the most human way possible?" — Amit PasupathyGuest: Amit PasupathyHost: Lanre Adeoye