Pure Signal

The episode contains a handful of practitioner-grade ideas - an agent registry for governance, a finops analogy for AI cost management, and the coordinator-vs-orchestrator distinction - but these are buried under extended hedging, meandering co-host monologues, and standard consulting platitudes about data foundations and change management.

“We're proposing also creating like, an agent registry that would have each agent kind of registered. We would know what they are doing, we know what their skill sets are, and we would have governance over the agents”

“just like cloud has financial ops and there's a lot of rules and systems in place and uh, wisdom and how to deploy that tool effectively. We will need to come up with a similar finops model for AI”

The episode largely recycles familiar consulting frameworks - people/process/technology, change management, don't skip steps - with only a couple of genuinely fresh angles: the insider detail about Adobe soliciting pricing model input from its advisory board, and the claim that hyperscalers are artificially suppressing true AI costs.

“right before summit I was in the um, the customer advisory board sessions and they actually asked the group I was in, how would you like us to potentially price these agents?”

“we are not really feeling the true cost of using these tools. Like I think they're artificially deflated as the scalars have gone out. And at some point there will become a, hey, the price to do this all the way up through this vertical integrated chain is going to be 10x 15x 20x”

Both guests are genuine Adobe ecosystem practitioners with real implementation experience in regulated industries and agentic AI architecture at a mid-sized consultancy, which is credible but not exceptional - neither has operated at hyper-scale nor carries a record of independently verifiable outcomes.

“I serve as a chief architect over Adobe content supply chain so designing uh, implementations of Adobe products like Workfront aem, um, integrations using fusion with various Adobe and third party platforms”

“I also uh lead our pre sales practice for um, the Omnicom Adobe practice, the new newly formed Omnicom Adobe practice”

The episode is almost entirely abstract, with only named tools (Workfront, AEM, Fusion, Claude) and a single unverified anecdote as concrete evidence; there are no client names, no hard metrics, no before-and-after data, and the one external example (Starbucks) was explicitly flagged as unconfirmed.

“the viral story of the week, which we'd have to go double check the source but you know, the whole Starbucks rolling out AI to go count milk jugs in the stores”

“one of the best implementations I ever worked on was they made a whole campaign around this technology implementation. They did a naming contest, they had logos submitted”

The host asks leading, multi-part questions and explicitly pre-softens them ('feel free to not use specific examples'); the co-host frequently hijacks turns with extended personal musings; there is no pushback, no challenged assertion, and no follow-up that extracts deeper specificity from either guest.

“Bethann, I would love to start with you just from your perspective. In the last three to five years, are you seeing what I. I'm seeing, or am I a little off base there”

“of just of all these things that we're talking about in your eyes, and feel free to not use specific examples, but just in general, what, what are some of the mistakes”