The Business of AI

Tom provides a well-structured framework of five failure modes in AI industrial projects and articulates practical approaches to pilot design and scaling. However, the insights are largely applied best practices rather than novel discovery - the core ideas (start with problems, align incentives, design for scale, build trust) are sensible but not unexpected to someone familiar with enterprise transformation or product-market fit. The value lies in contextualization to heavy industry rather than fundamental insight.

“the first is lack of operational ownership. Right? So a lot of the time currently AI pilots are sponsored by uh, innovation strategy or digital teams. Now whilst those teams are really important, they're not always the people who own the operational problem.”

“scale really has to be designed from day one. If you only start thinking about scale after the pilot, you're already too late.”

Tom applies the five-failure-modes framework and emphasizes design-for-scale and trust-through-transparency, which are sensible but largely in circulation within enterprise AI and digital transformation circles. The idea of drones + AI for asset inspection in remote terrain is tactically novel to his company but not a fresh conceptual breakthrough. The reframing of cost-cutting as value-creation is pragmatic but not counterintuitive. Missing are contrarian takes or first-principles challenges to conventional wisdom.

“It's not an AI pilot or a company that says, go on IT team, go and run me an AI pilot and come back to me and tell me what it can do for us. It's never going to really succeed”

“you scale by earning confidence. And that means transparency. It means human in the loop design, it means clear metrics and operational relevance.”

Tom Greenlees has legitimate operational pedigree: 20 years in consumer marketing at major brands (Coca-Cola, Red Bull, BAT), completed an executive diploma in AI at Oxford, co-founded a company in the space, and has worked directly with Permian Basin operators and oil/gas engineers. He is a founder-practitioner rather than a pure consultant or academic, which is valuable. However, he is not a C-suite operator at an established heavy-industry leader, and Intelligent Core appears to be an early-stage venture, limiting his demonstrated track record of scale.

“I spent 20 years of my career predominantly in the consumer landscape...working m for brands like Coca Cola, Red Bull and most recently British American Tobacco where I was global VP of Consumer experience”

“I went back to Oxford University, did an executive Diploma in Artificial Intelligence with a focus on heavy industry”

Tom provides limited concrete metrics or case studies. He names Coca-Cola, Red Bull, British American Tobacco, Exxon, and mentions the Permian basin and San Francisco driverless cars in passing, but offers almost no specific outcomes, dollar figures, or before/after metrics. The five-failure-modes framework is well-explained but lacks numerical support or named examples of projects that failed or succeeded due to these factors. The drone capabilities mentioned (four-hour flight, future thousands-of-kilometers range) are vague technical aspirations, not deployed results.

“We're developing a drone that uh, we, we anticipate can fly you know, thousands of kilometers, not just hundreds.”

“our long range drones...We're looking at our sensor packages as well and continually to innovate and drive development within how we collect data”

Tim asks reasonable open-ended questions and does invite Tom to elaborate, but largely allows Tom to monologue without pushing back, challenging vagueness, or requesting evidence. When Tom discusses drones, smart cities, or future products, Tim accepts broad statements without pressing for specifics (e.g., 'which Permian operators?', 'what scale of cost savings?'). The host does reframe and reflect (e.g., smart homes analogy), but rarely confronts claims. There are few productive disagreements or sharp follow-ups that would test Tom's reasoning.

“Yeah, uh, thank you Tom. That's a really good list.”

“That's really interesting to see. And I think uh, the thing you said um, around designing it as a demo, I think that is really key insight there”