AI-native services transform insurance operations and economics

The advertised accuracy was 95%. When Jamie Cuffe audited the actual work coming out of traditional BPOs, it was as low as 50%. He shared that at our AI-Native Services Summit in New York last week. The gap between what incumbent services claim and what they deliver is the opening. And closing it expands the market: Pace reviews 100% of claims in real time instead of sampling 1%, so work that used to be impossible becomes the standard. Jevons paradox at work in AINS. Some other takeaway from the AINS Summit: Where to start: if providing an external service to do something that is currently being done internally, find critical but not core tasks. Chuck Ganapathi is making the SaaS to AINS transition at Gainsight, a company with hundreds of millions in ARR. They started their renewals service on the long tail of accounts most companies are currently neglecting and have found that to be a great entry wedge. Trust is the constraint, not capability. Ryan Daniels made the argument of the day: people forgive human mistakes far more easily than AI mistakes. Ask Waymo. The human-agent split in AINS becomes a design problem: humans have to oversee the right work and own the errors. But AI needs to do an ever increasing amount of the job. The unlock ahead is using eval scores to get agents underwritten with real liability insurance. CC AIUC-1 What an eval looks like in production. Chris Hladczuk's agents at Hanover Park categorize cash transactions, and every human approval or correction becomes a labeled training example. He tracks the share of cash reconciled by agents, week over week. The payoff: day zero financials in an industry that runs 30 to 90 days behind. Eric Newcomer was in the room too. Today he published a piece asking whether AINS is the next SaaS, built around the playbook deck I presented. My answer to his headline is no. SaaS is a few hundred billion dollar market. Legal, accounting, insurance, and tax generate trillions a year. The next SaaS would be a disappointment. Eric's piece is in the first comment. 400 people were on the waitlist for this summit, so we'll continue to host more. There's a link to sign up for that as well. Thank you Ryan Daniels for hosting us at Crosby's stunning new office, and Chuck Ganapathi, Jamie Cuffe, Chris Hladczuk, and Carlotta "Lotti" Siniscalco for bringing it.

The gap between what traditional services promise and what they actually deliver is the opening. At our second AINS Summit, Jake Saper and a room of founders and operators dug into how AI-native services can close that gap—and fundamentally change the economics of the work. A few of the biggest lessons from the room ↓

Billing insurance is so much more than simply submitting a claim. Behind every payment are billers researching denials, reviewing EOBs, correcting claims, following up on aging accounts, verifying benefits, submitting documentation, handling pre-authorizations, and spending hours trying to resolve issues that often began on the insurance side. And the process seems to keep getting more complicated. Insurance companies continue to add requirements, change guidelines, request additional documentation, and create more administrative steps before a claim can be paid. What may look like a simple claim from the outside can require multiple follow-ups and hours of work behind the scenes. I don’t think people always realize how much responsibility falls on billing professionals. We are often the bridge between the provider, the patient, and the insurance company, trying to keep revenue moving while navigating systems and rules we didn’t create. Billing is not just “sending claims.” It’s research, problem-solving, persistence, communication, and a tremendous amount of follow-through. The work may happen behind the scenes, but it deserves to be seen.

Let’s settle the debate: Owner-Operator vs. Leased-On. I came across a post where both sides were flexing their muscles, debating who made the better business decision. The independent owner-operator says, “Why would I give up a percentage of my revenue?” The leased-on owner-operator fires back with, “I get better tire discounts, cheaper insurance, fuel programs, and back-office support.” Here’s the funny part: they’re both right. The leased-on truck may have access to better insurance rates, fuel programs, tire discounts, and other benefits because they’re part of a larger group. But those benefits aren't free. The tradeoff is typically a percentage of the revenue and, in some cases, less control. The independent owner-operator keeps more of the revenue—but also keeps more of the responsibility. When something goes wrong, you're handling it. Legal issue? You're finding the attorney. Compliance? That's on you. IFTA, 2290, insurance, safety, billing, collections, permits—the list goes on. You're not just the driver. You're also the compliance department, accounting department, safety department, and CEO. Meanwhile, someone leased to a larger carrier may have an entire back office making sure those things get handled. Neither business model automatically makes you smarter, more successful, or more profitable than the other. The better question isn't “Who keeps the bigger percentage?” It's “What are you getting—or giving up—for that percentage?” Pick the model that fits your lifestyle, your business goals, and the amount of responsibility you're willing to carry. Because sometimes keeping 100% means carrying 100% too.

The Insurance World May Slow Down for Summer – But RiskVille's Innovation Doesn't While the insurance industry may have slowed down for a few weeks, innovation doesn’t have to. At RiskVille, summer is far from a pause button. It's an opportunity to keep building, refining, and improving the platform our customers rely on every day. Continuous improvement, all year round Cloud software isn't something that should stand still. We believe a modern insurance platform should evolve continuously not just through major upgrades every few years. Throughout the summer, our development team has continued to deliver enhancements, performance improvements, usability refinements, and new functionality, ensuring our customers return from their holidays to a platform that's even better than when they left. Small improvements. Big impact. Not every improvement makes headlines, but many have a meaningful impact on day-to-day operations. Recently, we’ve strengthened our security controls and further enhanced the overall resilience of the system. Alongside this, we’ve continued to refine the user experience—improving UI options, introducing new notification features, expanding column choices and customization within report configurators, and extending our BI endpoints. These are just a few examples of the steady stream of enhancements delivered over the summer. Ready for what's next When autumn arrives and business activity picks up again, our customers don't return to the same platform they left behind—they return to one that's been continuously improved. Because while the insurance world may take a well-deserved summer break, innovation doesn't. And neither do we.

Growth creates a strange problem inside insurance agencies. The team gets busier. More policies are written. More clients call. More service requests come in. Revenue goes up. But so does the number of small things that can quietly fall through the cracks. A renewal follow-up gets delayed. A client calls twice because the first message never reached the right person. A producer finishes a conversation, but the notes never make it into the AMS. A service request sits in an inbox because everyone assumes someone else owns it. None of these problems look serious on their own. That’s why they’re easy to ignore. But at scale, they become the operating cost of growth. I think this is where agencies should look at automation first. Not at the biggest, most impressive AI project. Look for the small handoffs happening 50, 100, or 500 times every week. Then ask: What happens when one of these handoffs fails? That usually tells you where the best automation opportunity is. Because scaling an agency isn’t just about doing more business. It’s about making sure the systems behind the business can handle more too.

A few days ago, I turned in my key and my computer after 23 years as a financial advisor, and I handed $1,000,000 in revenue to 2 friends. We sat around a table and parceled out the clients. Two financial advisors I love and trust got a life-changing pay increase in a single afternoon.  For the last 2 years, I worked 1 day a month inside that practice, and I still managed nearly $1,500,000 in revenue on that schedule, just to prove it could be done. Most financial advisors are working 80, 90, 100 hours a week and are nowhere near $1,000,000 in revenue. They won't get to the 1 day a month with another designation, another certification, or another conference. They need a calendar built by design and a team that actually operates the company without you standing in the middle of it. One of the financial advisors I coach told me he could not take a vacation that week because he was personally negotiating his insurance contracts. That advisor is in our Club Deca program, which is a room for practices between $1,000,000 and $3,000,000 in revenue. He's running a strong practice on paper, but still negotiating insurance contracts. Busy is not his problem. He is too central, and an owner who is too central cannot sell, rest, or leave. So ask yourself the honest question. If you walked away for 30 days, would your practice quietly keep running, or would it quietly stop?

Insurance taught AI to read a claim in seconds. Then it mails a check. I sat down with Rayne Morgan at InsuranceNewsNet to talk about the one part of the stack nobody puts on the transformation roadmap — and it's the only part the policyholder actually feels. The payout. Here's the uncomfortable math: you can compress a 14-day adjudication into 14 minutes, and your customer still grades you on one thing — how fast the money showed up. Everything upstream of that is invisible to them. But speed isn't even the real problem. Legacy payment rails produce no data. A wire or a paper check tells you exactly one thing: a payment left the building. No claim number. No adjuster ID. No merchant. No category. No jurisdiction. Nothing a model can learn from. So the moment a beautifully instrumented AI workflow reaches the payout, your AI goes dark. That's not a payments problem. That's a transformation ceiling. What we build at Pliant is the opposite of that. Custom card programs rails that pay vendors, providers and claimants in real time with unlimited custom fields attached to every single transaction. Claim number. Policy/Member/Adjuster ID. Service category. Captured at the moment of payment, not reconciled back into it weeks later. Two things fall out of that, and they surprise most carriers: 1. Less risk, not more. You stop collecting bank details to pay someone. Less PII on file is less PII to breach. 2. Controls move upstream. Merchant category, amount, jurisdiction, time window; enforced before the money moves, instead of clawed back after. A warning I gave Rayne, and I'll repeat it here, because it's the thing that quietly kills the build: Before you modernize payouts, ask the vendor four questions. → Who owns the transaction data — me or you? → Can I export all of it, in a usable format, on any day I choose? → What happens to my data when the contract ends? → Can I get out of my contract with you whenever I desire without a penalty? None of this ships from a single department. It takes the CFO, claims, digital transformation, and IT moving at the same time. Every payout project I've seen stall, stalled because one of those four seats didn't have buy-in. Insurers have spent enormous energy making the front of the claim intelligent. The back of it still runs on rails built before the internet. That gap is where the next decade of advantage is sitting, and it's a lot easier to close than most carriers think. Where does AI go dark? I'd like to know👇 Thanks to Rayne M. for the sharp questions and a fair, well-reported piece and to the InsuranceNewsNet team for putting a spotlight on the least glamorous, most consequential part of the claim. Full article in the comments. repost♻️share #InsurTech #ClaimsManagement #InsurancePayments #DigitalTransformation #EmbeddedFinance #AIinInsurance #PropertyAndCasualty #Fintech #ClaimsInnovation #Payments

Premium-based licensing is the only vendor fee that goes up because you did your job well. You write more premium. The bill goes up. You did not add users, add compute, or ask for more support. You grew, and the vendor invoiced you for the growth. That is a tax, and it lands hardest in your best year. The line that finally hit scale. The program you doubled. The book you absorbed from a competitor exiting the market. Every one of those raises the vendor's take without raising what the vendor delivered. The invoice is not the real cost. The real cost is what it does to your decisions. A new program pencils out until you model the license step-up. A book you could take on becomes marginal after the vendor's share. You start sizing growth against a contract instead of against your appetite and your capital. Most carriers agreed to these terms when premium was smaller and the clause read like a footnote. It compounds quietly, then surfaces at renewal, when walking away means replacing the system that runs the business. 🇺🇸 Nick Lamparelli is moderating a session with Luke Magnan (CIO of Combined Ratio Solutions) and John Lucker (Universal Shield Insurance Group) on how this got engineered into core system contracts, the three places lock-in is written in, and the first move out that does not require rip-and-replace. September 16th Register here: https://luma.com/xqr0do1w

A newly launched insurance brokerage platform recently reported something notable: before making its first acquisition, it spent a year testing an "AI-first operating model" across ten agencies, and average profitability rose by more than 50%. That's a real, specific, publicly reported outcome. The company itself attributes it to "a combination of revenue and productivity gains" — not to AI alone. That phrasing is worth sitting with, because it's more careful than the headline version of this story usually gets credit for. A reported outcome tells you what happened. It doesn't tell you what produced it, or which conditions would need to remain true for it to happen again — and those are the questions that actually matter if this result is going to inform anything beyond itself. So I'd want to separate the pieces the company has already named. What changed on the revenue side, specifically? What changed on the productivity side? What role did the AI-enabled operating model play in each, versus what role was played by everything else that comes with introducing a new way of working — clearer process, closer implementation support, more attention on the work than usual? The company's own language leaves room for all of that, which is exactly why it's worth asking rather than assuming. None of this is a case against the result. It's a case for understanding it before treating it as a template. The company is now scaling this model across many more acquisitions than the original ten. Whatever combination of factors actually produced the 50% lift, some of those conditions will travel with the company as it grows, and some may not — and which is which matters enormously to whether the number repeats. That's the real question worth asking before copying a result like this: what actually produced it, and which of those conditions have to stay true for it to happen again?