AI and Data Horizons

AI and Data Horizons

By Michael GaleBusinessTechnology
Download on the App Store

All AI and Data Horizons episodes

  • Michael Stonebraker - Zero Percent: Why Enterprise AI Stalls at the Data Layer

    What happens when you point a large language model at a real enterprise data warehouse? Michael Stonebraker's team found out. Using MIT's warehouse, with 1,400 tables and real queries from the SQL log, the LLM scored zero percent. Not 80. Not 40. Zero. Add agentic AI, RAG, and every prompt-engineering trick available, and the model still got more than four out of five questions wrong. Meanwhile, public benchmarks report accuracy north of 80%. In this conversation on AI & Data Horizons, Stonebraker explains why those numbers mislead, what it means for any leader betting on AI inside the enterprise, and why the next few years will look very different from the hype. Stonebraker is a Turing Award winner who led the original POSTGRESproject at UC Berkeley.

    48 min
  • Robert Hodges - Real-Time Analytics and the Database Convergence

    Robert Hodges, CEO of Altinity and a database engineer for more than 40 years whose career runs from Sybase to Continuent, which he led through its sale to VMware, joins the podcast to explain why real-time analytics and transactional data are finally converging. He traces the split back to the mid-1990s, when analytic databases arrived with big pages and columnar storage, and land on agentic AI as the force pulling the two worlds together. They cover where the data comes from, from security systems flagging malware lookups within 20 seconds to LLM training runs and high-frequency trading, and why open source systems like Postgres, ClickHouse® software and Kafka have an edge because AI models learned on them. Robert also lays out a four-part architecture of transactional databases, analytics engines, event streams and data lakes, built to hold up however the AI revolution plays out.

    ClickHouse is a registered trademark of ClickHouse, Inc. https://clickhouse.com/

    23 min
  • Faisal Hoque - Building Structural Integrity for Agentic AI

    Faisal Hoque, founder of SHADOKA and NextChapter, and the #1 Wall Street Journal bestselling author of Transcend and Reinvent, joins the podcast to argue that the structural-integrity discipline long applied to hardware, software, and networks now has to extend to autonomous agents. Hoque points out that an agent can be spawned in ten minutes, which only raises the bar on what has to hold up underneath it — how data is defined, how it's governed, and whether the system it runs on is sovereign or somebody else's. Asked to design the "parachute" agentic AI needs, Hoque reframes the challenge as one of reversibility: never hand over a capability to govern something you can't wind back down, the same override authority commercial pilots retain on largely automated flights. He also offers a test for spotting which leaders will actually pull this off: in his experience, the ones who reach for sound bites and want an instant answer are "the losers," while the ones who succeed are willing to sit with harder, more complex questions.

    Faisal Hogue 2

    38 min
  • Hannah Hunt - Software as a Defense Asset

    Hannah Hunt discusses the use of software and its value as a fundamental asset in the defense of the country and its global standing. She highlights the significance of having soldiers who can code and build software to meet the challenges of the future operating environment, where warfare is expected to be more agile and occur in contested environments. Hannah's focus is on soldier-led software development, its potential impact, and the challenges and opportunities associated with this approach in the military.

    36 min
  • Bill Nussey - Energy Is Leaving Its Mainframe Era

    Bill Nussey, author of Freeing Energy and Partner at Tech Square Ventures, joins host Michael Gale to discuss how solar and batteries are technologies rather than fuels, and that this single distinction is the most underrated force in energy. Solar has fallen roughly 400-fold in price over four or five decades with batteries on the same trajectory. Unlike every energy system before it, this one scales down to a rooftop, a car, or a lantern—though in the US it costs nearly three times what it does in Australia, a gap he attributes to regulation rather than equipment. What makes the conversation especially relevant amid today's debate over AI's energy constraints is the bottleneck Nussey keeps returning to: how fast you can get generation switched on, not whether it exists.

    40 min
  • Shai Shen-Orr and David Harel - Precision Medicine And The Patient Experience

    Shai Shen-Orr and David Harel of Cytoreason talk about the future of precision medicine. The idea of being able to go to your personal doctor and asking how your immune system is doing seems like science fiction, but is likely to be a reality soon. Hear from two leading experts using AI and cell-based research to reveal the future of medicine.

    34 min
  • Rajnikant Gupta: The Decision Engine Behind AI That Scales

    Rajnikant Gupta, Global Head of Partner Ecosystems and Alliances at Tata Consultancy Services, argues that enterprise AI stalls not for lack of technology but for lack of a real problem to solve. The fix is what he calls a "decision engine." With host Michael Gale, he traces a shift that runs deeper than tooling: from data-aware to context-aware, reporting to reasoning, automation to autonomy, with AI as a decision coach and humans in the loop. Raj calls this an "intelligence choice architecture," and the enterprises that miss it are the ones whose AI struggles to scale. Listen to the full conversation to hear how TCS ran the same playbook on itself, generating 260,000 ideas and roughly 110,000 apps and agents in a single global hackathon.

    34 min
  • Tom Siebel - Retraining the Workforce for the AI Economy

    A conversation with Thomas M. Siebel, Founder, Chairman, and Chief Executive Officer of C3 AI, and Author of "Digital Transformation", on why the hardest part of AI was never building the technology—it's getting people to use it—and why reskilling and continuous learning are what separate the organizations that succeed from the ones that stall.

    34 min
  • Dustin Alpert and Frank Conway: Why the World Cup Is a Preview of Commerce to Come

    At a World Cup match, the clock isn't only measuring the game—it's measuring whether the business behind it can keep up. Dustin Alpert and Frank Conway of Shift4 join the podcast to explain why the world's most demanding venues become the testing ground for cutting-edge commerce, where tens of thousands of fans buy in one tight window and every second counts. They break down why frictionless, "invisible" commerce only works on a unified, real-time data foundation. When payments, point-of-sale, ticketing, and loyalty feed one system, an operator can act while the fan is still in the building; when they sit in silos, the moment is already gone. They also look at where things head next, from AI working behind the scenes and as a fan-facing concierge, to the rise of agentic commerce.

    22 min
  • Daniel Yarmoluk - The Context Window Bottleneck

    Daniel Yarmoluk, AI architect and founder of GraphifyMD who is also an adjunct professor at the University of St. Thomas, joins the podcast to discuss why the context window—not raw model power—has become the real bottleneck for enterprise AI. As token consumption soars and agent workforces scale, Daniel and host Michael Gale reframe the challenge as one of architecture: most context capacity is consumed by orchestration, rules, and probabilistic RAG, leaving only a sliver for the domain knowledge that drives good decisions. They explore what metrics CFOs should really be prioritizing, such as cost-per-inference and "intelligence per watt." They also examine why knowledge graphs and ontologies, not more pattern matching, are the way to reduce hallucination and drift, especially for high-stakes industries.

    22 min