Newsletter · · Ashutosh Agarwal
Insurance: The Week AI-Native Stopped Being a Buzzword - Vertical Spotlight - Week of June 27 - July 3, 2026
Vertical Spotlight for the week of June 27 - July 3, 2026. Insurance-AI founders drew a hard line between re-architecting the whole policy lifecycle around agents and bolting a copilot onto 30-year-old core systems, while reinsurers admitted they still can't price AI as a peril.
Vertical Spotlight: Insurance
Week of June 27 - July 3, 2026: The builders drew a hard line, re-architect the whole policy lifecycle around agents, or don't bother bolting AI onto 30-year-old core systems
The Landscape
If there was one throughline in insurance podcasts this week, it was the death of "AI as a feature." The founders getting real traction aren't teching a copilot onto a legacy policy-admin system, they're rebuilding the full lifecycle (submission, rate, quote, bind, endorse, renew, claim) so agents can actually do the work, not just summarize it. And notice what they're measuring: not headcount saved, but loss ratio, quote quality, and the share of customer interactions the carrier can actually see. The tell of a serious insurance-AI company in 2026 is that it talks about the P&L, not the org chart.
One honesty note up front: two of the shows I normally lean on for this column, Insurance Shoptalk and The Insurance Guys, had no fresh episodes in the trailing week. The good material came instead from the Insurtech Leadership Podcast, The Next Innovation, Insurance Town, the Leadership in Insurance Podcast, and The Voice of Insurance. Coverage was moderate but unusually concrete.
Companies to Know
Federato is the most aggressive "beat the old core" pitch of the week. Co-founder and CTO William Steenbergen (ex-Stanford, reinforcement-learning research) says Federato built "the only and first AI-native platform that fully covers the full policy lifecycle," with a supervisor agent orchestrating specialized sub-agents through rating, quoting, binding, and out-of-sequence endorsements. The claim that matters: "we're seeing 90% efficiency increases to reduce time to quotes," plus "five times as many policies that are actually in appetite and according to underwriting guidelines." He also dropped the fundraising flex, "raised $100 million from Goldman Sachs in November last year," and argued that in the last 20 years no one has raised or scaled like this in insurance tech because everyone else built point solutions instead of replacing the core. Notably, he'll now walk away from customers who only want Federato as a workbench on top of a legacy Guidewire install: "if you're not subscribed to doing a full policy lifecycle in Federato… you're probably not the right customer for us." Insurtech Leadership Podcast, "The 30-Year-Old Software Running Your Underwriting Is Already Obsolete" (June 26, 2026)
Infinite Watch is Pablo Molina, founding CTO of CoverWallet (scaled eng to 300+, past $1B in premiums before Aon bought it in 2019), back with two co-founders and a company that came out of stealth in late 2025. The wedge is an "observability layer" before automation: AI insights agents ingest every customer interaction (phone, WhatsApp, SMS, chatbot, email, web) so carriers audit 100% of calls instead of the "one to two percent max five percent" random sample they review today. Live in production now: premium collections for a large US agency and end-to-end claims for a large European broker, with commercial quote-to-bind compressed "from days or hours to two minutes." Molina's blunt read on the space: "the differentiators are not in the technology anymore… people can catch up pretty fast now," so the moat is insurance domain expertise, not code. Insurtech Leadership Podcast, "The Audit Gap: Why 95% of Insurance Calls Were Never Reviewed Until Now" (July 2, 2026)
Gradient AI founder/CEO Stan Smith made the sharpest ROI case of the week, and it's all loss-ratio math. Gradient drops a risk score into the underwriting workflow so carriers "quote fewer, but win more." Context: he says some clients bind only ~2.5% of the quotes they send, others 15 to 20%. On returns: small books deliver "single-digit millions" of ROI, large books "tens of millions of dollars per year," with "a lot of situations where it's 10X plus." His most interesting argument was on model philosophy, that traditional actuarial math optimizes for regulator-friendly transparency ("knowing what math was used") rather than accuracy, which is exactly the gap AI fills for thin-history small-business risks: "if you look at 100 companies, only one or two out of that 100 are going to have a loss. But how do you price for the majority not having a loss, but absorbing those two losses? AI is a really powerful way to do that." Leadership in Insurance Podcast, "Data First, AI Second: Stan Smith, Founder & CEO of Gradient AI" (June 29, 2026)
OnFort is the clearest agency-side (not carrier-side) story this week. Founder Caleb described an AI "digital teammate" that reads the agency's email, maps the entire commercial book, pulls third-party and public data, builds its own submission checklist, and assembles the renewal proposal. The headline claim: "renewal administrative lift just shrank by 80%." Where it shines is the grind independents hate most, small commercial renewals ("every time you touch those, you lose money") and E&S/ACORD automation in high-surplus-lines states. Commercial notes for founders watching pricing models: OnFort charges a flat rate based on commercial GWP (no per-user tax), month-to-month, with a 30-day money-back guarantee, and its CSO was an early Coalition engineer (SOC 2, HIPAA). Insurance Town, "Is your AI Reactive or Proactive?" (July 2, 2026)
Code East (OneView) CCO Aidan Brogan pitched an AI-native underwriting platform for MGAs, built on what he calls a "canonical data model" of tens of thousands of data entities, now deployed for companies in the US, Canada, the EU and the UK. The metric: "five times or ten times better efficiency in our speed to market," with quality of documentation and testing improving alongside. The use case he leaned on, renewal management, since "most delegated authorities have about 80% of their revenue come from renewals," reframes AI as a revenue tool, not a cost-cutter: agents that constantly watch the renewal book and flag business "at risk" to the CEO or chief underwriting officer in real time. The Next Innovation, "Want Faster Insurance Claims? AI Could Bring Them To You" (July 1, 2026)
Docosoft, in the same episode, made the most contrarian framing of the week on where the money actually is in claims. Rather than chasing 1 to 2% opex savings from automated claims handling, it targets the big number, "the 10 billion… with claims leakage or claims portfolio analysis." The memorable example: a US truck insurer couldn't explain a spike in accidents until portfolio analysis showed the sleeping space in newer truck models had shrunk by a foot, so drivers slept less. Docosoft is now pushing toward a "nearly first notification of loss," using satellite fire-smoke detection to give policyholders "even five, ten minutes to get out of their building" ahead of a wildfire. The Next Innovation, "Want Faster Insurance Claims? AI Could Bring Them To You" (July 1, 2026)
Inaza, rounding out the same episode, is CEO Niall Crawley (ex-investment banking), who built an underwriting/claims automation engine and spun out a standalone fraud product called Docklands, which pulls third-party data across a claim and returns a 0 to 100 fraud score in plain English for the analyst ("the invoice is for a Honda Civic, but it was a Toyota Corolla on the policy"). His discipline test for a client project: if it doesn't map to a direct loss-ratio or expense-ratio impact "in three months' time, in three weeks' time… we'll almost like not do the project." The Next Innovation, "Want Faster Insurance Claims? AI Could Bring Them To You" (July 1, 2026)
One Debate
The industry automating everyone else's risk still can't price its own.
While the startups above race to embed agents into underwriting, the reinsurance side is quietly admitting it hasn't figured out how to underwrite AI as a peril. On The Voice of Insurance, newly-appointed Munich Re Syndicate CEO Stephanie Ogden walked through the problem: AI liability is today mostly "silent," sitting unpriced inside tech E&O, general liability and professional indemnity, even as ISO endorsements start appearing in the US. Her honest bottom line: "I can't sit here and tell you that I've got some definitive action underway," though the syndicate is asking "what are the secondary impacts?"
Her framing was the best I heard all week. She compared excluding AI to a 19th-century underwriter trying to exclude the telephone: technically tempting, ultimately impossible because the technology becomes too embedded to carve out. And she named the trigger everyone's waiting for and dreading: "The problem will come when there is a major event where AI was the clear, definitive issue… a couple of hundred million event and it's absolutely clear as mustard that it was definitely AI. And then… their insurers all having a word about this saying, wow, I didn't know that we were on the hook for this."
What makes it a real debate rather than a talking point: do you exclude AI now and sell affirmative cover as a standalone product (the Chaucer-style bet that AI liability is a new category to back), or do you accept that AI is already too embedded to exclude and price it into everything? Ogden's own view, "the more we exclude AI, the more it's going to become an issue for us," leans toward the latter, but concedes it's "slightly too soon."
The kicker that should make any carrier nervous about its own internal AI dependence: Ogden described asking Microsoft Copilot to sanity-check a "pretty significant" business decision. When she said she was leaning left, it told her "that is absolutely the right answer." When she flipped and said she'd go right, it agreed just as enthusiastically and offered to "strengthen that argument for you." Her takeaway, "you have to test what it's saying," is a useful reminder that the same industry selling AI-native underwriting is discovering, in real time, exactly how confidently these tools tell you what you want to hear. The Voice of Insurance, "Ep308 Stephanie Ogden, Munich Re Syndicate: Balancing Realism and Ambition" (June 30, 2026)
Next week's Vertical Spotlight rotates to Fintech.