Today in Tech

Companies have spent billions racing to adopt artificial intelligence. But as AI moves from experimentation into production, are businesses discovering that the real costs are much higher than expected? In this episode of Today in Tech, host Keith Shaw talks with Erwan Menard, senior vice president of product at Crusoe, about the economics of AI, the hidden costs of AI agents, and whether the industry's massive investments in infrastructure are sustainable. The episode explores why GPU pricing and token costs tell only part of the story, how businesses should measure AI return on investment (ROI), and why inference workloads are changing the economics of enterprise AI. Keith and Erwan also discuss the growing role of open AI models, how companies can reduce costs by choosing the right model for each task, and why electricity availability may be a bigger obstacle to AI growth than GPU shortages or funding. Topics covered: * Why AI agents can dramatically increase inference costs * The hidden expenses of scaling AI from pilot projects to production * How businesses should measure AI ROI and successful task completion * Why open models and smaller language models could lower enterprise AI costs * The growing energy crisis facing AI data centers * Community backlash against data center construction and electricity demand * Whether companies are overbuilding AI infrastructure * Are we experiencing an AI bubble, or is demand just getting started? Menard argues that AI's biggest challenge isn't a lack of demand or investment. It's the ability to deliver enough power and infrastructure to support continued growth. But Keith challenges that optimism, questioning whether AI spending, circular investments and infrastructure expansion are creating an unsustainable boom. Is AI delivering enough value to justify its growing costs? And who ultimately pays for the infrastructure behind it?

AI agents are moving beyond simple chatbots and isolated tasks. The next step is agent-to-agent communication, where AI systems operated by different companies could negotiate, exchange information, make decisions and even conduct business with each other. But what happens when those agents misunderstand each other, hallucinate, get hijacked or make a costly decision without a human catching it? In this episode of Today in Tech, Keith Shaw talks with Swamy Kocherlakota, Chief AI Officer at Zscaler, about the promise and risks of agent-to-agent communication. They discuss how AI agents establish trust across company boundaries, why identity and access controls become critical, the danger of compounding hallucinations, where humans still need to stay in the loop, and why autonomous AI transactions will need the equivalent of “circuit breakers.” The bigger question: Is autonomous agent-to-agent business really around the corner, or are AI vendors getting ahead of their skis?

AI was supposed to save us time. Instead, it may be creating a new problem: AI slop (polished-looking emails, reports, presentations, research and code that someone else has to review, fact-check and fix). In this episode, Today in Tech host Keith Shaw talks with Chris Bevil of Commvault about the growing “review tax” of generative AI, why executives and employees may have very different views of AI productivity, and what happens when AI-generated information starts becoming accepted as corporate truth. They also dig into productivity theater, AI checking other AI, why people may trust AI more than it has earned, and a disturbing deepfake scam involving fake executives. Is AI actually eliminating work, or just moving the work to someone else?

AI tools such as ChatGPT, Claude and Copilot are spreading rapidly through the workplace, but simply giving employees access to AI isn't the same as teaching them how to use it effectively. In this episode of Today in Tech, Keith Shaw talks with Anthony Salcito, general manager of Enterprise at Coursera, about why companies are struggling to turn AI investments into meaningful productivity gains and why the missing piece may be human skills rather than more technology. They discuss why AI has created so much anxiety about job displacement, why companies can't treat AI deployment as employee training, and which skills workers actually need to succeed in an AI-driven workplace. They also explore the growing importance of critical thinking, decision-making, communication and lifelong learning, along with the role of certifications and microcredentials as jobs continue to change. Plus, Anthony explains what schools and universities are getting wrong about preparing students for AI and why banning students from using generative AI may miss a much bigger opportunity.

Hackers are using AI, social engineering, and legitimate business tools to make cyberattacks look like normal work. Instead of breaking through security systems, they study how employees communicate, impersonate trusted vendors, manipulate email conversations, and exploit everyday business processes. Piotrek Wojtyla of Abnormal Security joins Today in Tech to explain how attackers are improving phishing and business email compromise attacks with AI. He discusses fake invoices, vendor impersonation, compromised email accounts, shared inbox vulnerabilities, prompt injection, and the risks created when companies give AI agents access to sensitive workflows. The episode also examines why traditional security awareness training may no longer be enough, how cognitive overload makes employees more vulnerable, and what organizations can do to protect workers without expecting them to identify every sophisticated attack. Can companies continue trusting email, vendors, and automated systems when cyberattacks look almost identical to legitimate business requests? Subscribe to Today in Tech for weekly conversations about cybersecurity, artificial intelligence, enterprise technology, and the trends changing how we work.