Eye On A.I.

Companies are spending billions building AI factories, but most of them can't tell you why their AI workloads are failing, whether their GPUs are actually being used, or what their infrastructure is going to cost them when agents start running at scale. Paul Appleby, CEO of Virtana, joins Craig Smith to discuss the findings of their AI Factory Reality Check study, a research report that reveals a striking and underappreciated gap between the pace of AI infrastructure investment and the governance needed to run it safely and efficiently. Six in ten enterprises, the study found, cannot automatically identify root cause when an AI workload fails, a problem that compounds fast once you're running critical services on AI infrastructure at scale.
The conversation covers the mechanics of Virtana's observability platform, capturing 20,000 metrics per second across the entire AI stack, correlating them in real time, and increasingly using agentic capabilities to remediate failures automatically, but its most important insights are structural. Appleby makes a sharp observation that cuts through a lot of AI optimism: token costs are falling, but token consumption is exploding, meaning the total cost of running agentic AI systems is still going up even as the per-unit price drops. He also tracks a cultural shift inside enterprises - IT resilience reporting that used to happen annually now happens weekly - as evidence that technology risk has become a board-level conversation in a way it simply wasn't before. The result is a conversation that's less about the promise of AI and more about what it actually takes to make it work at production scale.
Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.

AI is moving faster than enterprise security systems were designed to handle. In this episode of Eye on A.I., Craig Smith speaks with Bradon Rogers, Chief Customer Officer at Island, Island about how companies are struggling to govern the rise of AI agents, browser-based workflows, and unsanctioned AI tools inside the workplace.
The conversation explores why traditional "block-and-control" security models are breaking down and how a new approach, embedding policy directly into the browser and user workflows, may offer a path forward. It also dives into emerging risks like prompt injection and autonomous agent behavior, and why enterprises are increasingly becoming multi-AI environments by default.
Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.

Most AI is built for people sitting at desks. Kriti Sharma builds it for the people who work in refineries, aircraft hangars, and utility networks responding to wildfires at 4 a.m. and she spends weekends on-site with them to make sure what she builds actually holds up. In this episode, Kriti joins Craig Smith to discuss what industrial AI really looks like when failure genuinely isn't an option, and why the gap between an impressive AI pilot and a production-grade AI system is so much wider in the physical world than most technology companies appreciate.
The conversation is grounded in three specific products from Nexus Black, the elite AI unit Kriti leads inside IFS. The first is Resolve, a predictive maintenance platform built in close collaboration with William Grant's - the distillery behind Glenfiddich and Hendricks Gin - that is projected to save £8.4 million per year at a single factory by reading complex engineering schematics, identifying failure patterns before they occur, and giving frontline technicians step-by-step guidance on their phones without requiring them to remove a safety glove to type. The second is an airworthiness compliance tool for commercial airlines that automates a process currently consuming weeks of human engineering time, where a single mistake carries regulatory fines of up to $20 million and grounding a fleet costs $140 million per day. The third is a disaster response coordination system for utilities, built in partnership with Anthropic, designed to help field crews coordinate during wildfires, hurricanes, and grid outages in ways that, as a California disaster responder told Kriti directly after the most recent wildfire season, will get communities back online and hospitals lit up faster than ever before.
Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.

What is an AI agent, really? Strip away the hype, and it's a model with access - to tools, APIs, databases, email, anything that lets it take real action instead of just generating text. That access is exactly where the risk lives, and Devvret Rishi, GM of AI at Rubrik, and former co-founder & CEO of Predibase, joins Craig Smith with a string of real-world incidents that make the case concrete: AWS reporting four major outages in 90 days after deploying coding agents, a Meta-related agent that deleted someone's emails while they were actively asking it to stop, and Rubrik's own internal pilot catching incidents that, without governance in place, would have gone unnoticed.
The conversation lays out the impossible choice most enterprises are facing right now - block AI agents and forfeit the ROI boards are demanding, or grant access and hope nothing breaks - and walks through how Rubrik's approach uses small, fine-tuned AI models to enforce plain-English security policies on every single agent action in real time. It closes on one of the most underexamined risks ahead: as agents increasingly talk to other agents to get work done, a layer of activity is forming that no human is watching, and the question of who's accountable when something goes wrong in that layer is only getting more urgent.
Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.

It costs up to $2 billion and fifteen years to develop a drug, and big pharma still fails half the time at the final stage. BullFrog AI founder, Chairman, and CEO Vin Singh joins Craig Smith with a clear diagnosis of why: the industry keeps picking the wrong drug target from the beginning, and no amount of downstream optimization fixes a fundamentally wrong starting point. Built on AI technology originally developed at Johns Hopkins' Applied Physics Lab, BullFrog has assembled a three-stage platform that cleans messy clinical data, runs causal analysis to map disease pathways, and then ranks competing drug targets using a competitive framework that removes the subjectivity most pharmaceutical decision-making still relies on.
The most striking results in this conversation come from two case studies: work with the Lieber Institute for Brain Development - analyzing thousands of post-mortem brains - that led to the identification of potential driver genes for depression, bipolar disorder, and schizophrenia in months from data that researchers had spent fifteen years studying, and a pancreatic cancer trial where BullFrog's platform identified a patient subgroup with survival rates three times higher than the study average. Vin also delivers a candid assessment of the broader AI-pharma landscape: more than 90% of AI deals in the space are missing their milestones, most companies are wrapping open-source tools rather than building genuine technology, and the shakeout between players and pretenders is already well underway.
Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.