FintechTalks

The episode contains some substantive information about the BSA/AML compliance market size ($300B globally, $60B in US) and the founder's experience at Revolute, but relies heavily on soft advice about relationships, trust, and founder mindset. There is minimal technical detail about how the AI actually works, what specific accuracy improvements were achieved, or concrete metrics on customer outcomes. Much of the conversation drifts into generic founder wisdom rather than actionable insights for B2B operators.

“there's a $300 billion sort of industry built around hiring you know humans to operate these very manual repetitive workflows. In the US alone, that's that's 60 billion”

“even for a fintech like Revolute, there were still, you know, tens of thousands of reviews done daily by a team of you know, four or five hundred”

The framing of deploying agentic AI to reduce false positives in AML compliance is sensible but not particularly novel - this is a well-trodden application area. The emphasis on relationships and trust-building as a go-to-market lever is standard startup playbook. The discussion of accuracy requirements (99.99%) is relevant but presented as straightforward necessity rather than generating fresh strategic insights. No contrarian claims or first-principles rethinking are evident.

“agentic AI to automate financial crime compliance operations, BSA AML within banks and financial institutions”

“if the bank in an FI can't trust the person, how can you trust the product, right?”

Rim Shah is a legitimate founder with relevant hands-on experience: ML background, time at Google, and lead of the financial crime product team at Revolute. He has direct insight into the problem space and has built products in this domain. However, he is still early-stage (under 2 years at Avre AI) and is not a seasoned operator with proven scale-up experience or a track record of building category-defining companies. He is solidly practitioner-level but not the caliber of a veteran battle-tested executive.

“So my background's in machine learning, uh so I have quite a technical background”

“spent a bit of time at Google, um, and then after that ended up going to Revolutes, yeah. Um, where I eventually ended up leading the financial crime product team”

While the episode names Revolute, Google, and Avre AI, and cites the $300B and $60B market figures, it lacks concrete operational metrics: no customer names (beyond 'large institutions'), no specific accuracy/false positive reduction numbers, no timeline data on customer implementation, and no revenue or growth metrics. The discussion of the 99.99% accuracy requirement is stated but not evidenced with examples of what happens when that threshold is breached or what competitors achieve. Advisors are named generically (ex-CRO of Wells Fargo, head of risk at US Bank) without specifics on their contributions.

“there's a $300 billion sort of industry”

“we're working with some very large institutions”

The host asks open-ended questions and does follow up on some responses (e.g., diving into how he convinces banks, probing the chip allocation question). However, the conversation often meanders without sharp follow-ups on key claims - e.g., when Rim mentions 'large institutions' and 'core banking partnerships,' the host doesn't press for names, deal sizes, or status. There is no pushback or productive challenge on claims; the dynamic is warm and affirmative rather than investigative. The host's questions are friendly but lack the incisiveness needed to extract concrete evidence.

“how do you convince them of that?”

“if you had a hundred chips, if you will, like how many, how much of how many of those chips would be on like technology”