AI Can Pay for Itself. Dr Alex Leathard Says the Hidden Costs Are the Ones Nobody Measures.

Every AI business case sells the same line: adopt it and it pays for itself. Faster work, lower cost, a return you can point to on a slide. But someone who has spent his career watching organisations actually try to make that maths work thinks the promise is the easy part and the part almost everyone gets wrong. The real costs of AI are the ones that never make it onto the spreadsheet: the culture that has to change, the value that's hard to measure, and the bill that arrives long after the pilot looked like a win.

This week on the Project Flux podcast, Dr Alex Leathard takes apart why so much AI adoption quietly stalls. It isn't the technology — it's that firms treat AI as a software rollout when it's really a change-management problem, and then measure the wrong things when they try to prove it worked.

🎧 Short on time? Hit play here — a grounded, hype-free conversation about why AI adoption breaks, how to actually measure its value, and what value-based pricing does to the economics.

Leathard's starting point is that adoption lives or dies on culture, not code. The organisations that get AI right treat a rollout the way a good scientist treats a claim: they run it as a hypothesis, test it, and are willing to be wrong. The ones that struggle buy a tool, mandate it, and wait for a return that never quite lands because nobody changed how the work actually gets done. The technology was never the bottleneck; the willingness to experiment was.

That leads straight into the measurement problem, which is where most ROI conversations fall apart. Some of AI's value is genuinely quantitative, hours saved, throughput up but a lot of it is qualitative: better decisions, less drudgery, people freed for work that's hard to price. Leathard's argument is that if you only count what's easy to count, you'll either kill a project that was working or keep funding one that wasn't. Measuring AI honestly means holding both kinds of value in view at once.

He's just as sharp on pricing. As the market shifts from per-seat licences toward value-based and outcome-based models you pay when the AI actually delivers a result the economics of adoption change underneath you. It's a better deal in principle, but it moves the risk around and rewards the firms that can prove what a result is worth. That ties back to his warning on the hidden costs: transparency, data provenance, environmental impact and bias aren't compliance boxes bolted on at the end, they're part of the true price of running AI responsibly, and they don't show up in the pilot.

Why it matters: AI adoption isn't broken because the models are weak, it's broken because firms treat it as a purchase instead of a change, and measure it with the wrong ruler. Run it as an experiment, count the qualitative value alongside the numbers, and price in the costs nobody likes to talk about. That's the difference between a return you can defend and a bill you didn't see coming.

👉 Listen to the full conversation with Dr Alex Leathard here.

Claude can be your analyst, editor, and strategist.

But most professionals are using it to fix grammar.

These 1,000+ Claude prompts take it from grammar tool to your most powerful AI work assistant.

Sign up for Superhuman AI and get:

Claim your free prompts

Enjoyed this? Project Flux breaks down the week's most important AI and tech news every Monday. Subscribe to the free newsletter here and get a step ahead on the world of AI.

Project Flux is independently owned, funded, and operated by its founder. Any external references or features are for community benefit only and do not imply endorsement, control, or ownership.