AI Radicals

Afleveringen

1. Everything Under the Model: What 7 Leaders Learned the Hard Way About Making AI Work

   4 november 2026 · 35 min

   Afspelen
   The hard part of AI isn't the model — it's everything underneath it and around it. Lees meer

   In this season 4 roundup episode of AI Radicals, host Satyen Sangani revisits seven conversations from the season to distill what it actually takes to make AI work inside a real company. The guests span the field: Charlene Li, co-author of Winning with AI; Brian Solis and Dave Wright, ServiceNow's innovation leaders; Erin McIntosh, who runs global data operations at CNA Insurance; Nathalie Berdat, data leader at the BBC; Anant Jhingran, CTO of IBM Software; Eugene Wu, Columbia professor and co-director of the Data Agents and Processes Lab; Francois Ajenstat, Founder and CEO of Golden Analytics; Mark Nelson, Venture Partner at Madrona; Julie De Moyer, who leads data and AI across fifteen beauty brands at LVMH; and Ben Taylor, Former Chief AI Strategist of Dataiku, who predicted most of this three years ago.

   Across the season, three things held. First, stop calling it a pilot — if it isn't strategic enough to go into production, it isn't strategic enough to do. Second, trust is earned one reconciled metric at a time, starting not with a problem but with a single number that two teams disagreed about. Third, the rules belong in the system, not in the prompt — a stop sign goes at the intersection, not in the head of the user. And underneath all three, something none of these leaders set out to say: the work is coupled, permeable, and never quite finished. Closer to biology than physics.

   "The risk of doing something is better than not doing anything in this space. And I really hate saying that, but that's really where we're at."

   Listen to this episode to learn:

   Why executives green-light AI pilots precisely because they haven't decided anything yet — and what changes when you frame AI as a strategic initiative instead

   How IKEA turned a customer service agent into a billion-euro interior design business by retraining 8,200 people instead of cutting them

   Where governance actually bel…

2. Open Source, Open Question: Why Model Supremacy Won't Last with Ashu Garg, General Partner at Foundation Capital

   21 oktober 2026 · 49 min

   Afspelen
   Why the most valuable asset in AI won't be the model — it'll be the layer that captures how your business actually makes decisions. Lees meer

   In this episode of AI Radicals, host Satyen Sangani sits down with Ashu Garg, General Partner at Foundation Capital and an early investor in one decacorn and six unicorns, to unpack where enduring value in AI will actually accrue. Four years ago, Ashu was non-consensus bullish on open source AI models. Today that view is consensus — and he's moved on to the next thesis: context graphs.

   Satyen and Ashu dig into the concept of context graphs — an infrastructure layer that captures the decision traces humans generate as they interact with AI applications, scattered across Slack, email, Salesforce, and a dozen other interfaces. They explore why the gap between open source and closed source models is shrinking to roughly six months, why 75% of tasks no longer need a better model, and what that means for the economics of foundation model providers.

   "The best AI applications will really need to capture decision traces over time and then figure out how to use those decision traces to improve the decisions they make in the future. That infrastructure layer is the context graph. And our belief is that the most valuable asset over time will be the context graph."

   Listen to this episode to learn:

   What a context graph is and why it can only be built in the context of a real application — not as a theoretical enterprise-wide exercise

   Why the six-month gap between open and closed source models means application developers must have open source in the mix, and why two-thirds of enterprise value will continue to be captured at the application layer

   When to fix problems in the model versus the context layer, and which incumbent application companies Ashu thinks are positioned to survive the AI transition — and which aren't

   --------

   “ Advancing your automation journey, it's a little bit like the levels of self-driving. As companies go through th…

3. The Control Plane Race: Who Governs the Agents with Victor Dey, AI Industry Analyst

   14 oktober 2026 · 49 min

   Afspelen
   Why the loudest narratives in AI rarely hold up once you talk to the people actually building the technology. Lees meer

   In this episode of AI Radicals, host Satyen Sangani talks with Victor Dey, AI and technology analyst contributing to Fast Company, Forbes, and the AI Journal, about what it takes to cut through the noise of enterprise AI — as a journalist who has interviewed the executives running nearly every major data and agent platform in real time.

   Victor and Satyen dig into how Victor's practitioner background in data science and NLP shapes the questions he asks CEOs, and why a practitioner's mindset increasingly matters for anyone selling, deploying, or building in AI. They unpack the story behind Victor's counter-narrative on the so-called "SaaS Apocalypse" — how one forthright conversation flipped the market narrative and sparked a wave of CEOs finally speaking up. They explore the race to own the AI "control plane," why competitors are simultaneously building and partnering on orchestration, and why context alone isn't enough: permissions, guardrails, and observability matter just as much.

   "What I think is more important than context, or equally important as context, is also permissions. What data did it touch? What points did it touch? Was it denied or was it told not to perform certain actions? I think that needs equal amount of light."

   Listen to this episode to learn:

   Why the SaaS Apocalypse narrative collapsed once a single CEO with conviction and customer evidence spoke up — and what that reveals about how market stories form and break

   Why the AI control plane is simultaneously the most contested and most collaborative space in enterprise tech, with competitors building rival products while partnering on the same capabilities

   Why context gets the headlines but permissions, guardrails, and observability deserve equal attention as agents proliferate and incidents mount

   --------

   “ With the rise of agents, I think this control plane about whether it be g…

4. Metadata to Knowledge Graphs: Mapping the Full AI Data Stack, with Sanjeev Mohan, Principal of SanjMo

   7 oktober 2026 · 45 min

   Afspelen
   Data foundations are the real bottleneck for agentic AI, and most teams are governing the wrong layer. Lees meer

   In this episode of AI Radicals, host Satyen Sangani talks with Sanjeev Mohan, industry analyst and author of Designing the AI-Driven Data Foundations, about why AI's biggest promises still rest on unglamorous data fundamentals, and why agents are breaking the old rules of governance.

   Sanjeev walks through his layered model of metadata, semantics, context, ontology, and knowledge graphs, and the two dig into where data products fit in that stack. They debate whether governance belongs to the agent or the data itself, why deterministic rules must live with the data rather than be re-explained to every model or prompt, and how agentic systems generate entirely new categories of data that nobody has a framework for governing yet.

   "There's a concept 100% related to the whole idea of context and data products, and that is governance. Now we are in the realm of AI, which is a very different beast to govern."

   Listen to this episode to learn:

   Why semantics, not storage or compute, is becoming the real moat in the AI stack

   Why agentic governance has to cover input, output, and runtime, not just clean data in and clean data out

   Why data teams are entering a golden era, shifting from backup-and-restore work to business strategy and reusable data products

   --------

   “ One definition is that data product is a philosophy. It's not a thing. And what data product says is that when you build your artifact, you build it with the product management best practices. For example, there's a data product owner, it has a version number, and it's backward compatible. It's packaged with SLAs and contracts. So if you are packaging your semantic layer and you're building it using the product management best practices, then that's a data product, very much so.” – Sanjeev Mohan

   --------

   Time Stamps

   [01:37] Sanjeev’s book Designing the AI-Driven Data Foundations

   [03:47] Defining metada…

5. Data Products, Not Platforms: A CDO's Playbook for the Agentic Era with Cara Tice, CDO of Early Warning

   30 september 2026 · 46 min

   Afspelen
   Why "America's data steward" is more than a job title, and what it takes to keep fraud intelligence trustworthy in the age of AI agents. Lees meer

   In this episode of AI Radicals, guest host Susan Wilson, Sales Leader at Alation, talks with Cara Tice, Chief Data Officer at Early Warning (the company behind Zelle), about what data stewardship really means when trillions of dollars in real-time payments, and the fraud intelligence protecting them, run on your data.

   Cara and Susan dig into why data products have replaced the "one data platform to rule them all" mentality, why integration and change management matter more than any single tool, and why AI doesn't create data problems, it just exposes the ones that were already there.

   "Your AI governance is reliant on your data management program. Make sure you've got data governance up and running, because AI is going to rely on your data management controls and the context you create."

   Listen to this episode to learn:

   Why "data products" — not centralized platforms — are the fastest path to trustworthy, reusable data

   Why AI governance is really just an extension of data governance, and why the two leaders need to be joined at the hip

   What a new CDO should fix (and what to leave alone) in their first 90 days on the job

   --------

   “As a CDO or as a data steward, you need to be obsessed with the customer outcome. Don't lose sight of that. It's easy to lose sight of it when you are focused on plumbing and fixing this issue and that issue. You always have to remind yourself of what's the bigger picture, what's the north star? With data products, you become relentless about the customer outcome because you're delivering to a specific business need, you're creating more capability that is business centric around data. When I think about the features that you would need for data science models, it's the same. Whether your consumer is a data scientist or is a commercial consumer, that data is used to make a decision. So you need to…

6. Semantic Coupling & the Interconnected AI Stack with Eugene Wu, Professor at Columbia University

   23 september 2026 · 55 min

   Afspelen
   Why AI agents fail at scale, and the systems-thinking fix nobody's building yet. Lees meer

   In this episode of AI Radicals, host Satyen Sangani talks with Eugene Wu, Columbia professor and co-founder of the Data Agents and Processes (DAPLab), about why building reliable AI agents is fundamentally a systems problem, not a model problem.

   Eugene explains "semantic coupling," the idea that every layer of an agent's stack, from data retrieval to tool calls to reasoning, is interdependent, so a small failure anywhere can quietly corrupt the final output. He and Satyen draw a parallel to early relational databases absorbing the complexity that applications used to handle themselves, arguing that today's computing infrastructure needs to do the same for agents: managing data flows, enforcing rules deterministically instead of hoping a prompt is followed, and giving agents safe room to explore and fail without real-world consequences. Eugene shares research from his lab on search at massive scale, why even top models struggle to find the right evidence in huge datasets, and a "branchable" computing environment that lets agents try, fail, and roll back cheaply.

   "You need somewhere to ground reliability and quality. If it's not probabilistic, and you can rely on it and it's guaranteed, then the system doesn't need to think about it at all."

   Listen to this episode to learn:

   Why "semantic coupling" makes traditional layered software thinking break down for AI agents

   Why pushing rules into the system (not the prompt) turns probabilistic behavior into guaranteed safety

   Why compute-level innovations like branchable environments could redefine what we expect agents to do on our behalf

   --------

   “ So many people in the data community and companies are working on data search, but how do you evaluate the end-to-end quality? And how much search is even the critical bottleneck in this kind of end-to-end question? Because what does the agent need to do? It needs to take your question, it ne…

7. The 90-Day AI Roadmap with Charlene Li, Author

   16 september 2026 · 42 min

   Afspelen
   Why AI transformations fail and the 90-day plan that actually works. Lees meer

   In this episode of AI Radicals, host Satyen Sangani sits down with Charlene Li, bestselling author and strategist, to unpack her new book on how organizations can create real value with AI, not just deploy it.

   Charlene explains why the biggest mistake leaders make is treating AI as a separate strategy instead of a tool that serves their existing business strategy, and lays out the 90-day framework she and co-author Katia Welch built to help executives move from confusion to a clear AI roadmap. They dig into why "Goldilocks governance" beats both reckless and overly restrictive AI policies, why ownership of AI shouldn't default to IT, and why the obsession with pilots is really a symptom of leaders abdicating strategic responsibility. Charlene also shares real examples from a call center that grew headcount instead of cutting it, to a bank that reskilled instead of laying off, showing what an abundance mindset toward AI looks like in practice.

   "Good governance doesn't slow you down. It actually speeds you up, because you know what you are able to do and what you shouldn't be doing."

   Listen to this episode to learn:

   Why AI should never be treated as its own strategy and how to anchor every AI decision in existing business goals

   Why "moving from pilot to production" is really a leadership gap, not a technology gap

   Why an abundance mindset is what separates organizations that win with AI from those that stall out

   --------

   “ They're green-lighting these pilots because they feel they need to be doing something, but they won't green-light into production because they don't have an overarching AI roadmap that supports their strategy. They haven't put the time and attention to it. They've given it to IT. They have not really thought about what are we trying to do with this? The business leaders have abdicated their responsibility for AI because they themselves don't understand it, and they're not …

8. The Case for Federation in the Age of Agents with Anant Jhingran, CTO of IBM Software

   9 september 2026 · 53 min

   Afspelen
   Why the same old data infrastructure playbook won't survive contact with agentic AI, and what actually has to change underneath. Lees meer

   In this episode of AI Radicals, host Satyen Sangani talks with IBM Software CTO Anant Jhingran about why enterprise AI's biggest bottleneck isn't the models, it's the decades-old data and integration infrastructure sitting underneath them.

   Anant and Satyen dig into why data fundamentals (provenance, metadata, systems of record) haven't changed, even as humans and fixed workflows give way to agents reasoning on the fly. They contrast "AI for data" with "data for AI," why AI's tolerance for messiness still doesn't excuse bad data, and why centralization may matter less as agents run quick, discovery-driven queries instead of big fixed reports. Anant also shares a new focus at IBM: rethinking whether one "golden" code path per product still makes sense when AI makes forking and personalizing variants easy.

   "If you think that agents are just going to do the same thing that you're doing except machines instead of people, it's kind of boring, and I don't think that's going to happen. The real change is they're doing something different that we haven’t done before."

   Listen to this episode to learn:

   Why "AI for data" and "data for AI" are two distinct problems enterprises need to solve separately

   Why agentic, discovery-driven workloads may reduce the need to centralize all your data, but raise the stakes on metadata

   Why forking products into many tailored variants, instead of one shared code path, could reshape how software gets built

   --------

   “ Something that I wouldn't have thought of three months back or six months back, which is how do we actually build products. And the reason is very simple, is that if you just say that AI is going to help us build products faster, then it doesn't actually create a competitive differentiation because everybody else is creating products faster with AI. So you have to both do things differently and p…

9. Infinite: Why AI Business Reinvention Beats Automation with ServiceNow’s Brian Solis & Dave Wright

   2 september 2026 · 51 min

   Afspelen
   Why mode one thinking keeps most companies stuck—and what it takes to build a company that can keep reinventing itself with AI. Lees meer

   In this episode of AI Radicals, host Satyen Sangani talks with ServiceNow’s Brian Solis and Dave Wright and authors of Infinite, about why so many enterprises get stuck chasing ROI on isolated AI use cases instead of using AI to become something genuinely new.

   Brian and Dave unpack their "mode one, mode two" framework: deciding what existing work deserves to scale with AI (mode one) versus using AI to unlock entirely new value the business couldn't create before (mode two). Using stories like Ford's costly rehiring of quality engineers after over-automating, and IKEA's Billie bot freeing thousands of agents to launch a billion-euro design business, they explain how the real ROI conversation starts with strategy, not use cases. They also dig into why most companies are still stuck optimizing yesterday's workflows, why trust and psychological safety are prerequisites for innovation, and why AI governance has to evolve from a checkbox exercise into managing AI as a true enterprise asset.

   "AI is not the strategy. If it does become the strategy, it very much limits the impact it's going to have on the organization."

   Listen to this episode to learn:

   Why leading with use cases limits AI's impact, and how IKEA turned 8,200 agents into a billion-euro business

   Why most companies stay stuck optimizing yesterday's workflows instead of reinventing them

   Why governing AI as an asset is key as agentic AI scales

   --------

   “ You'll see a common set of challenges, like, for example, what's the ROI of AI? That seems to be a popular conversation that has all kinds of different schools of thought around it. AI is not the strategy. If it does become the strategy, it very much limits the impact it's going to have on the organization and how you can measure its success. Where we have the more successful ROI conversations is if we take a step back and look …

10. AI Governance in Public Media with Nathalie Berdat, Data Director of Product at the BBC

    26 augustus 2026 · 46 min

    Afspelen
    How data trust breaks—and how to rebuild it before AI makes it worse. Lees meer

    In this episode of AI Radicals, host Satyen Sangani sits down with Nathalie Berdat, Data Director of Product at the BBC, to explore how one of the world's most trusted media institutions is rebuilding its data foundations for the AI era.

    Nathalie shares how she diagnosed a quiet trust crisis inside the BBC—teams producing conflicting numbers for the same metrics—and led a multi-year effort to fix it: identifying the handful of metrics that actually mattered, building certified "data products" as single sources of truth, and modernizing a legacy platform to support them at scale. She also unpacks why AI governance at a public institution carries different stakes than at a commercial company, how the BBC decides where genAI is (and isn't) allowed to touch editorial content, and what has to be true before agentic AI can responsibly run across an organization like the BBC.

    "The governance isn't a compliance checkbox, it's closer to editorial standards. It has to be defensible to a journalist."

    Listen to this episode to learn:

    Why low trust in data often shows up as two teams presenting two different numbers for the same metric and how to fix it

    Why the BBC treats AI governance as an editorial issue, especially when it comes to recommendations and content curation

    Why agentic AI requires clear data ownership, documented lineage, and machine-readable governance before it can be deployed responsibly

    --------

    “ Building a data product that gives you a very trusted source of truth when it comes to who works and where and what cost center allows you to then expose this product and build on top something like return on investment for our content or program, because then you'll know who has worked, how much it cost us to build and develop a program. You need to know your return on investment for something you'll be commissioning. You'll be investing a lot of effort and time and people on it.” – Nath…

Alle 89 afleveringen →