In Conversation With Ux Magazine

Everyone wants to talk about agents and models. Jeff McMillan, starts where almost nobody else does: the foundation.In this episode, Jeff McMillan, founder of McMillanAI, former Head of Firmwide AI at Morgan Stanley, and advisor on enterprise AI, maps AI as a stack: high-quality accessible data → semantic layer (knowledge graphs, RAG) → control and governance → models → orchestration → applications. The heavy lifting is in the bottom layers. Organizations that skip them can fake it for a handful of agents, but at 150 or 15,000 agents, you need near-100% accessibility and 99%-plus quality, or you’re monitoring chaos you can’t see.Josh and Robb press him on why knowledge management feels unfundable, why tribal institutional knowledge breaks when machines execute without judgment, and why evaluation (golden datasets, custom org evals, regression when models upgrade) is the work builders hate and operators can’t skip. Robb names the trap CTOs are falling into: grinding tokens on feature backlogs that never reach

Most organizations treat a bigger context window like a cheat code: dump every document in, skip the data work, ship. Jonathan Frankle, Chief AI Scientist at Databricks, says that's still wrong.This is Jonathan's return visit to Invisible Machines — a conversation recorded last summer, released ahead of Databricks Data + AI Summit. His first appearance (season 2) was the MosaicML-era craft conversation: lottery tickets, mixology, mini-cupcakes. This one is the enterprise engineering thread: be a scientist, curate before you scale, and treat specification (what you actually want the system to do) as the bottleneck between raw model power and useful AI.Robb and Josh press him on the myths that still seduce enterprise teams: million-token windows as a substitute for real data work, hyperscaler résumés as a proxy for talent, and the fantasy that unlocking every PDF in the org automatically makes knowledge useful. Jonathan's answer is consistent: measure success, test your use case, climb the ladder of techniques,

Mastercard's central AI team receives roughly a thousand requests a year from across the organization. A few years ago, most of them were for chatbots. Today, most are for AI agents. Federico Cohen Freue, Executive Vice President of AI & Data Operations at Mastercard, has watched this shift in real time and knows exactly what it reveals about how enterprises are (and aren't) thinking about AI.In this episode, Federico explains why the name people use for what they want matters less than whether they understand the conditions that make it work. “Ball bearings,” as Robb Wilson puts it: demos can't reveal the difference between a solution that will hold and one that will blow up the engine. What actually matters is training, fluency, and a clear framework for where to deploy AI with purpose.For Mastercard, that framework is deliberate: use AI to make commerce more secure, smarter, more personal, and to make the company itself stronger. Not everything. Those things. The simplicity is a feature, it gives a spr

Most AI strategies are just a buying plan: literacy workshop → vendor shortlist → adoption scoreboard. Brian Evergreen (Founder of The Future Solving Company, author of Autonomous Transformation) argues that this sequence explains a lot of failure, and it isn’t strategy at all.In this episode, Brian reframes the job: set the technology aside long enough to name the new value you want to exist, in language vivid enough that people can feel the outcome. From there, no strategy without vision: you work backward through “what would have to be true,” turning invisible opinions into a visible map of bets before agents, data estates, or org charts get to pretend they’re the point. According to Brian, “10% more profitable” isn’t a vision, and a moonshot can still be concrete.Josh and Robb press him on the pressure to remove friction and flatten the middle of the org. Brian doesn’t dismiss friction work, he warns that friction can quickly pile up if you go hunting without a north star. Vision is the force with enough

In season one of Shell Game, Evan Ratliff sent a voice AI version of himself out into the world. In season two, he launched a startup staffed entirely by AI agents. What he ended up with was a live experiment in what these systems actually do and what they do to us.Each of the agents working for Hurumo has a name, a role, a personality, and an expanding, though usually unreliable, memory. Kyle the CEO became a character people either loved or hated. A version of Megan from marketing turned up in a Hertz hold queue. The whole project was a side door into what's actually happening when AI systems are given a job and set loose.In this episode, Evan joins Josh and Robb to go deeper on what he learned. On the very human complexity of what a job actually is and why "this person does skill X, AI can do skill X, therefore AI can replace this person" is a fundamental misreading of how organizations work. They explore how generative hallucination isn't just "getting things wrong" — we've built the most successful confa

Most organizations treat crisis as a failure state. Marina Nitze sees it as a window.Nitze served as Chief Technology Officer of the Department of Veterans Affairs (the largest civilian agency in the country) during the healthcare.gov collapse. She helped rescue it, helped stand up the US Digital Service, and came out the other side with a question she and her colleagues have been pursuing ever since: why is it that crisis makes otherwise impossible transformational change possible?That question became a firm, Layer Aleph, and now a book, Crisis Engineering, co-authored with her colleagues. In this conversation, she walks through what a "useful crisis" actually looks like, the five indicators that distinguish it from chronic problems masquerading as crises, and the practitioner toolkit for standing up a crisis engineering center when the window opens, because the window is usually hours, not days.We also get into two stories that hit harder than any framework: the California unemployment system's call center

There's a book about artificial intelligence that doesn't start with Sam Altman. It doesn't start with Elon Musk. It starts in 1994, at Cambridge, where a teenager named Demis Hassabis is reading Gödel, Escher, Bach and concluding, before most of his professors would have agreed, that first-order logic can't be the full answer to building intelligence.Sebastian Mallaby spent years inside that story. His new book, The Infinity Machine: Demis Hassabis, DeepMind, and the Quest for Superintelligence, is the most serious attempt yet to explain not just what AI is, but why the people building it can't stop. His answer draws on a line Jeff Hinton borrowed from Robert Oppenheimer: invention is sweet. A scientist, given the chance to build something, simply cannot resist. The consequences come later.In this conversation, Mallaby joins Josh Tyson and Robb Wilson to explore the full sweep of the Demis Hassabis story — from game designer to neuroscientist to Nobel laureate to the man running Google's flagship AI lab. The

The IRS has roughly 60,000 fax machines, and nobody can get rid of them. Not because there’s a law that says you have to use them (there almost certainly isn’t), but because likely decades ago a memo got written, somebody interpreted fax machines as the most secure transmission method, and that memo calcified into what Jennifer Pahlka calls "folk law," a perceived rule that nobody can locate, nobody can challenge, and everybody treats as immutable.Folk law looms large in the American government right now. Cascades of rigidity built from outdated interpretations of rules that were flexible to begin with, administered by people who were never asked whether any of it was working. Jennifer Pahlka, who wrote Recoding America: Why Government Is Failing in the Digital Age and How We Can Do Better, is the founder and former executive director of Code for America, and was Deputy CTO for Government Innovation in the Obama White House. She’s working on the gap between what government is supposed to do and what it actual

Most organizations are still implementing AI as point solutions, dropping new technology into existing workflows to do the same work, just slightly better. The real value lies in system solutions that completely transform how organizations operate. Avi Goldfarb, economist and co-author of Prediction Machines, joins Robb and Josh to explain why AI adoption follows predictable economic principles and why internal resistance, not technology limitations, is the primary barrier to transformation.This conversation, recorded back in 2023, reminds us that most organizations continue to struggle with the same issues surrounding systemic change in 2026. Goldfarb's core argument: AI is fundamentally cheap prediction. Just as the internet made search and copying cheap, AI makes prediction cheap. When something becomes a commodity, the complements, the things that work alongside it, become more valuable. This includes compute power (benefiting Microsoft, Amazon, Google), unique data, and crucially, human judgment.The prob

Joshua Gans, economist and co-author of Prediction Machines (and holder of the Skoll Chair in Technical Innovation and Entrepreneurship at the Rotman School of Management, University of Toronto) joins Robb and Josh to reframe how enterprise leaders should think about AI. Rather than chasing the hype around artificial intelligence, Gans argues we should understand AI as an advance in computational statistics that drops the cost of prediction, reduces decision-making friction, and fundamentally reshapes organizational structure.Many organizations are full of people waiting for phones to ring, managing buffers, absorbing uncertainty. As AI makes prediction cheap, this middle-management friction layer flattens. His new book, The Microeconomics of Artificial Intelligence, examines the ways AI enhances and perhaps enables decision-making, and how that’s poised to affect organizations and industries. The trio discusses the "hidden secret" of AI adoption that the people who choose the systems used to automate work ar

Before enterprises can deploy AI agents that actually work, they need something most organizations don't have: a single, authoritative source of truth. Joe DosSantos, Workday’s VP of Enterprise Data and Analytics, joins Robb and Josh for a wide-ranging conversation about canonical knowledge, the semantic layer, and why data governance, a concept from the 1990s, has suddenly become essential for AI deployment.Large language models are predictive engines modeled to anticipate what users probably likely mean. For B2C applications where multiple interpretations are acceptable, this works fine. But enterprises need deterministic truth, not probabilistic guesses. The trio outline a solution in three layers: establishing canonical knowledge, building a semantic layer to translate between human definitions and machine-readable formats like YAML, and using LLMs as an interface to deterministic back-end systems.For leaders evaluating AI investments, this episode clarifies what actually needs to be built before agents c

Ben Goertzel, the researcher who helped popularize the terms "AGI" and “singularity”, as one of the most influential modern champions and systematizers of AGI, returns to Invisible Machines to discuss the decentralization of AI and what's actually missing from today's most advanced systems with Robb Wilson and Josh Tyson.As enterprises rush to deploy AI agents and LLMs reshape workflows, a critical question emerges: who controls the infrastructure? Goertzel argues that while big tech dominates model development, a tension is building between centralized hegemony and decentralized, open systems — the same dynamic that shaped the internet itself.In this wide-ranging conversation, Goertzel discusses his current work on Hyperon (the successor to OpenCog) and the ASI Chain, systems designed to enable decentralized AGI development. He explains why the rapid cycles of AI hype and disappointment — the traditional "AI winters and summers" — no longer slow progress the way they once did. The speed of change has acceler

We’re in a moment when organizations are approaching agentic AI backwards, chasing flashy use cases instead of building the scaffolding that makes AI agents actually work at scale. Erika Flowers, who led NASA’s AI Readiness Initiative and has advised Meta, Google, Netflix, and Intuit, joins Robb and Josh for a frank and funny conversation about what's broken in enterprise AI adoption. She dismantles the myth of the "big sexy AI use case" and explains why most AI projects fail before they start. The trio makes the case that we're entering a post-software world, whether organizations are ready or not. Listen and learn why the scaffolding— or agent runtime — matters more than use cases, why organizational gaps kill AI projects, how to move projects from pilot to production, and what "post-software" actually means for enterprises.Check out Erika’s podcast, “Flower Power Hour”: https://open.spotify.com/show/15BTSl9fWiH3QTmVAYj6FdLearn more about Erika at www.helloerikaflowers.com/0:09 - NASA AI Readiness Explained

As 2025 draws to a close, Robb and Josh look back on some of the conversations they had this year both on the podcast and advising major enterprises and government leaders to offer their predictions for agentic AI in 2026. With major disruptive forces like outbound AI in the hands of consumers and agent runtime environments allowing organizations to create scalable infrastructure for AI agents, next year could see seismic changes in the way investors look at companies, and the ways companies look at themselves. Featuring a look at the components of an agent runtime, as well as previews of upcoming episodes with returning guest Ben Goertzel of SingularityNET and Joshua Gans, co-author of Prediction Machines, this episode is required viewing for anyone charged with finding ROI with agentic AI. 00:00 – Introduction to 2026 Agentic AI Predictions01:12 – Outbound AI Arrives02:30 – Scaling vs. Inventing AI04:55 – Ben Goertzel Preview06:45 – Scrappy Innovation in AI08:20 – Invisible Work Explained10:00 – Agents

How did a Dutch company most people haven’t heard of come to hold the fate of AI? Marc Hijink, author of Focus — The ASML Way joins Josh and Robb to explore the insatiably precise process of producing the chips that power GPUs, and with them the AI taking shape all around us. Marc is a financial reporter and technology columnist for the Dutch daily newspaper NRC, and Focus is the result of decades of uncompromising, embedded reporting on ASML, which produces 90% of all chips worldwide. Operating out of a quiet town in the Netherlands, their extreme ultraviolet (EUV) lithography machines produce a steady stream of chips working with an accuracy of within a few atoms. The trio explores how the EUV process is as much probabilistic as it is deterministic, as well as the impact that different cultures have on ASML’s partnerships and pipeline. Relying on lenses that have to be grown from crystals, lithography is a high-stakes endeavor that requires a delicate balance of tooling and advanced engineering. We n

Omar Santos is a Distinguished Engineer directing AI Security at Cisco. He’s here for a frank conversation about the realities of security in the agentic era. As more software is created on-the-fly by AI agents at the request of humans, security has to become an ever-present layer. Security will be built into complete agent runtime environments and will require constant human oversight and intervention, augmented by the ability to simulate outcomes to avoid risk. Omar is also the Co-Chair of the Coalition for Secure AI, and these are the things he’s thinking about on a daily basis. He sits down with Robb and Josh at the end of a travel blitz that included work surrounding OpenAI’s Stargate Project, a four-year $500b plan for new AI infrastructure in the United States. The trio discuss how the ongoing training of models and the rising demand for inference continue to push the demand for security across burgeoning technology ecosystems. ---------- Support our show by supporting our sponsors!This episo

As the CEO and Co-founder of Apollo GraphQL, Matt DeBergalis has a lot to say about the ways that organizations can confront their inherent complexity and build reliable systems for AI to flourish. He joins Robb and Josh to talk about how GraphQL has made it easier for developers to build meaningful, AI-powered solutions.As an open-source language for APIs, GraphQL makes data fetching more precise and flexible. When utilized within an agent runtime environment, GraphQL gives organizations the ability to build their own tools. As Matt explains, this gives orgs an enormous edge with agentic AI, and helping teams move light on their feet, experiment, and adjust quickly.The trio also reminisces about the early days of the internet and agrees that ChatGPT was the mother of all demos.Looking for the Agent runtime episode - https://youtu.be/CddjTUWSaHA?si=86dSwjyE3uZGYY8FHere is recent news Apollo announced at Summit last week :https://www.apollographql.com/newsroom/press-releases/apollo-expands-platform-to-power-th

Josh and Robb get together to share a short but deep dive into agent runtime environments. Agent runtimes are the missing piece in the many AI projects out in the world that never reach production. Robb explains how runtimes allow businesses to orchestrate AI agents so they can collaborate around business objectives and drive real ROI. Runtimes also give agents access to canonical “source-of-truth” knowledge bases and tools like MCP and A2A that allow them to pull the levers of existing software. Robb and Josh discuss the major investments in time and money required to build a runtime from scratch and why finding a complete agent runtime lets organizations make quick strides with the complex technologies associated with AI.---------- Support our show by supporting our sponsors!This episode is supported through partnership with OneReach.ai. Experience the first complete runtime for AI agents - a turnkey private architecture and tools for building enterprise-grade agentic solutions. Use any AI modelsCreate guar

Brian Solis, is a digital anthropologist and futurist who serves as the Head of Global Innovation at ServiceNow and is the former Head of Global Innovation at Salesforce. Brian is a bestselling author of Lifescale, The End of Business as We Know It, and Mindshift, which helps leaders learn how to see emerging trends, harness disruptive forces, and use them to fuel growth. Brian joins Robb and Josh to talk with us about the importance of self-disruption in AI adoption. With so many genAI projects failing to reach production, this episode looks at approaches taken by IKEA and Airbnb that have induced the kind of mind shift that invites self-disruption. Brian, who's been called the CEO whisperer, also talks about the nuanced difference between “aha” and “uh-oh” moments and the pitfalls of hopeless optimism.Brian Solis’s works: https://briansolis.com/books/Key Chapters0:00 — Intro: Josh & Rob set the stage0:34 — Guest intro: Brian Solis (ServiceNow futurist, author of Mindshift)1:07 — Why AI adoption req

As Global Head of AI at PayPal, Prakhar Mehrotra is looking deeply at the ways agentic AI is affecting commerce — particularly timely given PayPal and Google’s recent announcement of a multiyear strategic partnership to advance various commerce solutions. He joins Josh and Robb for a layered conversation that speculates on the shape of future marketplaces, where machines are bidding and setting prices on behalf of individuals and businesses. The trio also discuss the ways that AI is reworking the plumbing of technology within organizations and how agent runtime environments hold the key to successful adoption of AI. As the world evolves into one with more AI agents than websites, digital interactions might become simplified for users in ways that add complexity for businesses. Prakahr shares his insights into how humans might interact with technology in the future and how technology will change commerce in ways that are hard to fathom but crucial to confront.---------- Support our show by supporting our