Moody’s Talks: Risk Reframed

The episode contains a handful of genuinely useful practical ideas - the natural-language compliance playbook, confidence scoring on enriched data, and token-cost optimisation - but large stretches are basic AI-101 explainer content mixed with office-move chat and birthday-poem anecdotes. The insight-per-minute ratio is low for a B2B practitioner who already understands agents.

“when we give you an alert because of a new sanction, what do you do with it? And he actually shared his playbook for what his human analysts do”

“his remark to me was that with that he would actually do due diligence on a lot more of the customers and he would do it more frequently than he can today”

The framing is almost entirely standard industry narrative circa 2025 - agents as LLM plus harness, garbage-in/garbage-out, meet customers where they are. The most original moment is the concrete illustration of a CCO changing a geographic parameter in natural language, but the episode makes no contrarian or first-principles arguments.

“the chief compliance officer can go in there and change 50 miles to 75 miles. And see how that impacts his portfolio of customers”

“it's really not about sprinkling AI on top of the solution, but it's really embedding it where it makes sense”

Both guests are genuine internal practitioners who built and shipped the described agents, and Keith's story of co-designing the screening agent with a crypto firm's CCO demonstrates real operator-level experience. The limitation is that both are Moody's employees on a Moody's company podcast, which constrains candour and eliminates external validation.

“I went back to our very, uh, early kind of machine learning AI team and said, could we automate this? And we kind of partnered with that organization to actually build a pretty early, to be honest, agent that took that playbook and automated it”

“we've kind of red teamed some of this and tried to, to to create data in that way and really found that it's not practical”

There are a few solid data points - 90% analyst headcount reduction, 10% residual team size, three-to-five days for enhanced due diligence, the 1 Canada Square lookup demo - but no customer names, no revenue or cost figures, and most claims about competitor limitations or data quality remain asserted rather than evidenced.

“they offshored their analyst team and it's now 10% of the size that it was”

“three to five days on average, partly because there's a cost involved in doing that deeper due diligence”

The host draws out the crypto-firm origin story and the banking-conference compliance officer reaction, which are the episode's best moments, and he routes in audience questions. However, no claims are challenged, the opening third is heavy with small talk, and questions are largely leading and promotional rather than probing.

“You mentioned the story of the crypto provider Kepler. What was the impact on that in the end once it went into deployment?”

“how have you sort of thought through, when do we want to direct the builders in the Org to go about pursuing agents versus where we specifically don't want them to?”