Reading time: 5 minutes
Listen time: 6 minutes
​

Read the insurtech headlines this quarter and the story sounds simple: the money came back.

It did, $1.6 billion of it. But it is not chasing insurance brands anymore; it is buying the rails underneath them. And nearly every dollar is going to a handful of AI-native plays writing the exact coverage your institution sells.

The part the funding charts miss is the bigger shift: AI is now on both sides of your transaction, pricing the risk on one side and faking the inputs on the other.

This week I spent a day at Insurtech Trails, in a room with twenty insurtech founders and the VCs backing them. Whether you carry this coverage on your balance sheet or you are building the stack, the same shift lands on you. This brief is the macro. Wednesday, we take it to the floor.

We place the tech-heavy specialty risks the best underwriters want to write but need to understand first. Three things every carrier, MGA, founder, and board should have on the table before the next renewal:

• Funding rebounded to $1.6 billion, spread across the fewest deals since 2016. The capital is buying the rails under your business, and you will run on them whether you build them or rent them.
• A model can now read a submission, price it, and assemble the quote before a human opens the file. That convenience just moved your E&O somewhere your wording never looked.
• 57% of consumers have used AI editing tools, and 99% of insurers have already seen a fabricated document. The stack you plug into is only as honest as the inputs you can authenticate.

Every week our team rips through 200+ insurance, legal, regulatory, and market-risk articles so you don't have to.

This week, the arc that connects all three: where the money goes, where the model acts, and where it breaks if no one governs what flows through it.

Prefer the audio this week? Press play here: Listen to this week's Brief

The capital came back. It is buying infrastructure.

Summary

Insurtech funding rebounded to $1.6 billion in the first quarter of 2026.

Read past the headline and the shape is unusual. Deal count fell to 81, the lowest since 2016, while the median check hit $10 million, roughly double the 2021 peak. Capital is concentrating in fewer, larger, AI-native companies, and the AI label now dominates almost every dollar raised.

The clearest proof is Corgi, an AI-native carrier that underwrites startup D&O, E&O, cyber, EPLI, fiduciary, and AI liability in minutes, no broker in the middle. It raised at a $630 million valuation in January and crossed $2.6 billion by late May.

So what?

Follow the pattern, not the round.

Money is funding the rails that sit under carriers and MGAs, not the consumer brands that defined the first insurtech wave. In June, Portage Mutual, founded in 1884, handed its home and auto pricing to an outside AI engine, work that used to live inside the actuarial department. The same month, Feathery shipped a tool that mines a carrier's entire submission history to show why it wins and loses business.

For a founder, the infrastructure you would have built is now something you buy. For a carrier, those dependencies ride on your balance sheet, and AM Best is already asking how you govern them.

Underwriters have stopped asking whether you use AI; they want to know which parts of your stack you own, which you rent, and who answers when a rented model is wrong.

Source: (CB Insights, State of Insurtech Q1 2026; FinanceX, InsurTech's AI Takeover; placements via Insurance Innovation Reporter and BusinessWire)

Agentic AI is now handling the submission

Summary

The pitch used to be that AI would help underwriters work faster.

The product that shipped this year does more than help. Agentic systems now ingest broker submissions, enrich them with outside data, flag the gaps, and assemble the decision package before a human ever opens the file. In distribution, the same pattern runs through intake, triage, appetite matching, and quote orchestration. The work that used to sit with a person now sits with software that acts, and the underwriter across the table knows it.

The line between a tool and an agent is whether a human still has to press the button. With agentic systems, no one does.

The LION Lens

What happened — AI agents have moved into live distribution and underwriting workflows, handling submission intake, data enrichment, gap detection, and quote preparation that human staff performed a year ago (FinanceX, InsurTech's AI Takeover).

Why it matters — When a model triages a submission or matches appetite without a human in the loop, the professional judgment your E&O policy assumed a person was making is now made by software your wording never contemplated.

Practical implications — The failure most likely to generate a claim is the one your current tower is least built to answer: an agent that misroutes, misprices, or misrepresents, with no named human owner on the decision.

So what?

Agentic distribution relocates risk without anyone signing off on the move.

When an autonomous workflow binds the wrong appetite or feeds a broker a quote built on bad data, the error path runs from the model to the client to the carrier, and your contract becomes the first exhibit. The same delegation question lands on the carrier side at renewal, because the D&O underwriter now prices whether your board can explain where models act alone. Most insurtech wordings still read as though a person performed the service, which leaves a gap between what the policy imagined and what the software actually does.

The sharper risk is the blanket AI exclusion that carves an agent-driven loss out of the policy, the way some carriers already carve AI out of D&O.

The LION POV

Here's how we're advising clients:

• Build a one-page decision map before your next submission. One row for every distribution or underwriting decision, showing whether a human, an assisted model, or an autonomous agent makes the call.
• Name an owner for every automated decision. Not "the data team." A person accountable when the agent is wrong, with a gate that kicks edge cases back to them.
• Pressure-test your E&O and D&O wording against agentic failure now, while it is a negotiation and not a claim. The cheapest time to find the gap is before one finds you.

A clean governance story is what keeps an agent-driven loss covered, not just affordable.

Source: (FinanceX, InsurTech's AI Takeover)

Want to know how your distribution stack and tech E&O hold up before an underwriter asks? Contact LION Specialty for a confidential review.

Plug into the stack, then prove the inputs are real

Summary

The same generative tools drawing nearly every venture dollar are also in every claimant's pocket.

57 percent of consumers have already used AI editing tools, and 99 percent of insurers say they have encountered manipulated or AI-altered documentation. Insurers feel reasonably confident catching edits to a real photo. Only 32 percent feel confident spotting a deepfake.

An MGA that plugs into an AI-native submission and claims stack inherits speed and a new exposure at the same time, because the inputs flowing through it may be synthetic.

So what?

For a specialty MGA, plugging into the stack is the easy part.

Authenticating what flows through it is the work underwriters now probe at placement. They want to know how you confirm an input is genuine, and who is accountable when a fabricated document clears your automated review. Two-thirds of insurers already believe digital media fraud goes undetected often, and only 27 percent rate cross-carrier sharing as strong. That gap is a loss-ratio question for the carrier and an E&O question for the MGA whose model approved the claim.

The institutions that can show an input-authentication control and a named owner will clear underwriting ahead of the ones that automated speed without governing trust.

Source: (Verisk, State of Insurance Fraud Study)

The Bottom Line

The capital, the agents, and the fraud all converged on one place: the infrastructure layer where insurance decisions now get made. The institution that owns its stack, governs where its models act, and can prove what flows through them sets its own terms. Everyone on the other side of the table faces the mirror image of that question, and it belongs on the next risk committee agenda. Which rails do we depend on, where do our models decide alone, and can we show an examiner the inputs are real?

This was the macro. Wednesday, we break it down.

The conversation in that room never stayed up where the funding charts live. It went straight to the questions every operator lives with, like getting the numbers right, scaling a sales motion that began with the founder, managing a board, and reading what VCs actually underwrite before they wire. Friday maps the terrain. Starting Wednesday, the Intelligence series takes it to the founder and operator level, beginning with what is really driving that conversation.

That's why we created the D&O Contract Vigilance Blueprint. It's a 5-day email course to help you:

• Secure better D&O insurance: learn how to avoid common policy mistakes
• Protect your personal assets: understand your potential liability

Don't wait until a claim hits to find out your institution is under-protected.

Thank you for reading today's edition!

Want to share this edition via text, email or social media? Simply copy-and-paste the link below:

https://lionspecialty.kit.com/posts/insurtech-as-infrastructure-agentic-ai-in-distribution-and-underwriting-mgas-plugging-into-ai-native-stacks

And if this briefing was forwarded to you, subscribe directly here.

Stay Covered Out There Y'all,

FLIP

Co-Founder and Managing Partner
LION Specialty

P.S. Boards get exposed by D&O gaps they never see coming, and agentic AI just widened the gap. We built a 5-day email course on what to watch for. Comment BLUEPRINT and we'll send it.

P.S.S. Nothing in this briefing is legal advice; these are the opinions of the founder. It's market intelligence built to help you ask sharper questions of your advisors and make better decisions at your next insurance renewal.
​
​Thanks for reading! If you loved it, consider telling your colleagues (and/or) friends to subscribe.

If you didn't enjoy the email you can unsubscribe here.