The Question Every Underwriter Actually Needs Answered: What Does This Business Do?

Alan Ringvald founded Relativity6 in 2016, in grad school, with a co-founder and a genuinely nerdy starting question: everyone in business talks about customer lifetime value, but how accurately is anyone actually calculating it? The company built a machine learning platform to predict LTV with more precision, found early traction across several industries, and eventually landed on insurance — where lifetime value shows up concretely as two things carriers obsess over: a customer’s propensity to churn, and their propensity to buy an additional product.

Working across carriers, brokers, and wholesalers on that churn-and-cross-sell platform, Alan’s team kept running into the same upstream problem, in every single engagement: most of these organizations simply don’t know what their commercial customers actually do. It sounds trivial — a bank is a bank, a barbershop is a barbershop — but accurately classifying a business at scale, down to a six-digit NAICS code, turns out to be genuinely hard, and full of hidden risk signals a human reviewer would easily miss. Classification had quietly been the secret sauce underneath Relativity6’s LTV platform the whole time. Roughly two months before this recording, the team made the call to strip that classification engine out, turn it into its own standalone API, and build the company around it — a pivot Alan described candidly as painful (a full reset of the team’s mission and customer base) but clearly right, validated by 95 carrier and platform pilots in the two months since.

About Alan Ringvald

Alan Ringvald is Founder and CEO of Relativity6, an AI platform that predicts a business’s six-digit NAICS classification codes in real time, built specifically to serve commercial underwriters.

Why Real-Time Beats a Static Database

Relativity6’s key technical differentiator, per Alan, is that it isn’t a lookup against a static database — every API call runs a live machine learning inference, returning a result in under a second (averaging roughly 800 milliseconds) at 87-91% accuracy. That distinction matters because small and mid-sized businesses genuinely change what they do; a classification pulled from a database refreshed every few years can already be stale by the time an underwriter relies on it, and keeping a static database current at real-world scale is, in Alan’s words, its own unsustainable maintenance burden.

He described sitting directly with underwriters to watch the current process in action: a submission comes in, often missing key fields, and the underwriter has to manually research the business, validate what was submitted, and reconcile null values — a process that might take ten seconds for a simple dentist’s office or five minutes for something more complex like a financial advisory firm. Relativity6 doesn’t replace that judgment, but it removes the initial research step entirely.

Complementary to Ask Kodiak, Not Competing With It

Asked directly how Relativity6 differs from Ask Kodiak, Alan was unambiguous in his admiration — calling it a company he’s a huge fan of, technically and as a brand — and framed Relativity6 as a natural, complementary step that happens earlier in the same workflow rather than a competitor. Ask Kodiak focuses on matching a classified risk to the right underwriting appetite; Relativity6 solves the step before that — accurately determining what a business actually does in the first place, including secondary classifications a simple label would miss (his example: a barbershop that also serves alcohol to customers, a detail that meaningfully changes the risk profile even though “barbershop” is technically an accurate description). At the time of the conversation, the two companies were actively in discussions about working together.

Who Buys, and Why It Matters at the Point of Submission

Alan’s clearest priority customer is a carrier’s commercial underwriting group, followed by underwriting-focused platforms, then broker platforms, and finally brokers themselves — roughly in order of how urgently classification accuracy affects their business. He pointed to Coterie Insurance as a concrete customer example: because Relativity6’s API responds fast enough to fire within a submission form itself, it functions as a real-time validation step at the exact point a business describes itself — which matters especially for platforms doing direct-to-consumer, online underwriting, where speed is inseparable from the customer experience, and every extra second of waiting compounds into real user frustration.

The Hidden Cost Behind “Clean Data”

Alan offered a candid aside about machine learning in production generally: for all the language around generalized AI and deep learning, a huge share of real effort at large enterprises — Relativity6 included — goes into unglamorous data cleaning, since most insurance carriers simply don’t have clean, ML-ready data to begin with. He noted that even Google, despite its scale and infrastructure, deals with the same underlying problem internally. Relativity6’s pivot effectively converts that hard-won internal data-cleaning capability into the product itself — carriers no longer need to hand over years of raw transaction data (a process that used to take Relativity6 up to a year to negotiate through procurement); now it’s simply an API token they can test directly.

Advice: The White Lotus

For his closing recommendation, Alan kept it light: HBO’s The White Lotus, which he’d just started watching and found genuinely funny — an easy, low-commitment recommendation for anyone looking for something entertaining rather than heavy.

Key Takeaways

• Relativity6’s pivot from lifetime-value prediction to standalone business classification came from recognizing that classification had quietly been the hardest, most valuable part of its original platform all along — validated by 95 carrier and platform pilots within two months of the pivot
• The company’s core differentiator is running real-time machine learning inference (under a second, 87-91% accurate) rather than querying a static database — a deliberate bet that small business classifications go stale faster than most carriers’ existing tools can track
• Relativity6 positions itself as complementary to, not competing with, Ask Kodiak — solving the “what does this business actually do” step that has to happen before an appetite-matching tool like Ask Kodiak can work accurately
• Coterie Insurance uses Relativity6’s API directly inside its submission flow, since the response speed is fast enough to validate a business’s classification in real time without disrupting a direct-to-consumer underwriting experience
• Even well-resourced machine learning companies spend enormous effort on unglamorous data cleaning before any model can run reliably — a reality Alan noted applies just as much to Google internally as to an early-stage insurtech startup