IT Tech Videos & Podcasts

Every CEO Is Living Through Y2K again Overnight, Neil Fox argues, a developer can become a supervisor of AI agents. That is not a tooling upgrade. It is an identity shift, and most organisations have no plan for it. Neil, Chief Transformation Officer at Artizent, has spent a lifetime in software and now works as much with business leaders as with engineers. He explains why agile teams of seven to nine are shrinking to pods of three, why the bottleneck has moved from writing code to deciding what is worth building, and why treating AI as a time-saver misses its real value: running thousands of experiments at once. We also get into the token budgets CFOs have no idea how to plan, and the question Neil wishes more leaders asked: what does all this do to us as a society? Press play if your job title is about to mean something different.

Your Agents are Junior Staff. Who's checking their work? At one point, Jenny Larsson was letting AI do so much of her writing that she felt herself getting worse at it, until the words no longer sounded like her. So she pulled back. That small retreat sits at the centre of a much bigger problem. Executives talk about AI constantly, she says, yet almost entirely as a productivity play, while customer growth goes undiscussed and entry-level roles quietly disappear. If fewer juniors climb the ranks, who becomes the expert capable of judging an agent’s work? Jenny, former global CIO for specialty lines at Intact Insurance and now an adviser to executive teams placing their AI bets, unpacks her human-AI delegation ladder, research suggesting 70% of AI success is process, people and change, and why your visibility inside AI answers is the new search problem. Press play before you hand over another decision you can no longer check. On today’s show we also talk about “I love it, but I hate it”; a piece of research from Nuffield Health and the Ada Lovelace Institute into the impact of digital on young people. Read the report .

Is Your AI Pilot Too Snazzy to Survive Production? What happens when any employee can launch an agent that updates a thousand records before IT even knows it exists? Recorded during Dreamforce week in San Francisco, David Savage sits down with Frank Klensch, who leads technical architecture and engineering for Slalom's enterprise applications team, after more than a decade in and around Salesforce. With IT budgets under intense scrutiny and bottom-up AI experimentation spreading through organisations, Frank argues that the companies getting real value aren't the ones with the flashiest pilots. They're the ones that treat autonomy as something an agent earns.

Is Rewarding AI Spend The Dumbest Idea Ever? Last year, an IT leader told his C-suite the only thing he could guarantee about their new AI tooling was that they'd change their minds within six months. It took three. Now it's budget season again, and boards are demanding AI savings yesterday. Jim Daniel, Senior Vice President of Enterprise IT and Business Operations at Smarsh, argues that pressure is exactly why so many enterprises are deploying AI without governing it, and why so much of the output is slop rather than value. At Dreamforce, we unpack why the model is no longer the limiting factor, how a switch to usage-based pricing changed who gets access, why vendors quietly switching on AI is the governance gap nobody budgets for, and why he's relieved he no longer owns security. If your AI spend is climbing faster than your confidence in what it returns, press play.

AI Won't Fix The Hole In Your Roof A customer told Taksina Eammano something that morning that should worry anyone running a physical workforce: 17% of their people will retire in the next few years, and decades of know-how will walk out with them. Her bet runs against the doom narrative. While AI disrupts desk jobs, she expects a surge back into the trades, because no model is fixing a roof or rewiring a building, and the global data-centre boom needs people on site. Recorded at Dreamforce, David Savage and Salesforce's EVP & GM of Field Service unpack where the "human premium" is worth paying for, why robots are taking the dull, dirty and dangerous work first, and the question almost nobody is asking: how do you performance manage a team when AI makes everyone 70% as good?

Why I Spent $30k on a GPU and Lost Four Kilos Testing AI Three months after ChatGPT launched, Navan co-founder and CTO Ilan Twig bought a $30,000 GPU, spent months testing uncensored models and lost four kilos in the process. His question: if a model gives five different answers to the same prompt, how do you trust it in production? One finding stuck. It knows what a lie is, and will use one to avoid losing. That discovery shaped an AI agent that now resolves more than 62% of customer chats with zero critical hallucinations. We unpack why he's moving most of his AI onto small in-house models, why he calls the industry's reliance on Chinese open-weight models a ticking bomb, and why he believes the race to superintelligence cannot be stopped. Then comes the uncomfortable part: in that future, we're the gorillas.

Is Your AI Strategy a Science Project with a budget? Most AI programmes begin with the exciting part. Prashant Hinge, Chief Information and Transformation Officer at MSIG USA, argues that is exactly why so many aren't delivering value. He describes how the specialty insurer took a different route, the first transformation of his career to start with data: consolidating 15 insurance systems into a single data model in 12 months before building on top of it. We hear why he warned leadership the work could slow the business for the first 18 months, why a wrong number on a dashboard usually has to be fixed in the source system rather than the reporting layer, and why he believes the industry is jumping straight to AI while data, process and integration are still missing. He also makes the case that most basic process problems do not need a frontier model at all.

Rethinking Who Belongs in Tech Too often, the story of human trafficking ends at the rescue. The survivor leaders Laura Hackney worked with wanted something different: a job, stability, a career. So who gets to build the technology the rest of us use? Laura is co-founder and CEO of AnnieCannons, a US nonprofit that trains survivors of human trafficking to become software and AI engineers. Students join a 12-month paid programme with health benefits and work on production code while they learn. Many don't arrive wanting to be coders. They arrive wanting a career that supports their family. David and Laura discuss why financial independence is a critical part of the healing journey, and where the "rescue" narrative falls short. The biases companies bring about who "makes it" in tech, and how investor attitudes have shifted. How apprenticeships at growth-stage companies are converting at around 80% into full-time hires.

What to Fix Before Deploying Autonomous AI How do you secure autonomous AI agents when zero-trust fundamentals are still lacking? Mandy Andress, CISO at Elastic, joins David Savage to break down the reality of securing AI-driven enterprise data. From Elastic’s "Customer Zero" framework to grounding AI in context to prevent hallucinations, Mandy shares why tightly scoping permissions and deploying deception strategies are essential for modern resilience.