The Business of Digital Podcast

There are occasional genuine nuggets - the AI traffic vs. conversion inversion stat, the emotional prompt-improvement claim - but these are scattered across long stretches of high-level platitudes, meandering panel talk, and general statements about 'speed' and 'fundamentals' that add little for a practitioner.

“only 1 to 2% of the traffic is actually coming from AI. But something like 50 to 60% of the conversions. Are coming from AI.”

“if you tell the AI how important this is to you, it will actually improve the outcome”

The episode leans heavily on circulating AI-adoption narratives; the Ozempic speed analogy and the 'posers' categorisation from Kawaja's presentation are mild bright spots, but most arguments - privacy first, data as currency, human touch - are well-worn industry talking points.

“Ozempic is so popular because people want to get rid of fat quickly and they're willing to pay for it versus go to the gym for eight months.”

“there was a third bucket that he called posers. And those were companies that like have a. They say they do AI but in theory that it's just like machine learning”

All three guests are genuine domain practitioners - a data/programmatic CEO with 25 years of experience, a VP at one of Canada's largest broadcasters, and a Samsung legal counsel specialising in AI and privacy - but none are globally prominent operators and their contributions reflect mid-market Canadian-market experience rather than at-scale transformation.

“we've partnered up with the likes of Baneras, MasterCard and so forth. So in Canada specifically we see about half of transactions in the country”

“I've been in the advertising industry for 25 years now”

A handful of concrete data points (45,000 audience behaviours, the 1-2%/50-60% conversion split, the PromptMetheus platform, N8N vs. Make) give the episode some texture, but key claims are poorly attributed ('there is a stat that, you know, is out there') and the workflow descriptions, while directionally useful, rarely include timelines, dollar figures, or verifiable metrics.

“there is a stat that, you know, is out there that just in general, on average, um, there is, you know, implementation happening. And it more often than not, it's not giving people the outcomes”

“we're holding a database of 45,000 different audience behaviors”

The host occasionally surfaces an interesting angle (pushing on the 85% failure rate, asking about change management at scale) but consistently lets claims pass without challenge, poses broad open-ended questions, and openly hedges his own data ('you can check me on the data'), producing a friendly panel format rather than a probing interview.

“I, I think in one of our conversations you did mention that a huge number of AI, uh, implementations would fail at around a rate of 85%. You know, you can check me on the data.”

“Can you tell us what that journey has been like?”