Insurance software this week: three stories that matter (Nov 2–9, 2026)

This week, two distribution platforms opened their systems to AI agents built by other companies. Gallagher Re’s Q3 numbers showed that almost all insurtech money now goes to AI. And Vertafore added another AI agent for MGAs that comes with a large productivity claim. Here is what happened and why we think each one matters.


1. Your next user interface is someone else’s AI agent

What happened: On November 5, Zywave released the second part of its Fall 2026 launch. Its Producer Agent now runs through the company’s Apex MCP server, so producers can use it from general-purpose assistants such as Claude and ChatGPT. Producers still approve every email before it is sent. The next day, Lemonade said on its Q3 earnings call that it plans to open Lemonade OS to outside AI agents. (PR Newswire, Yahoo Finance)

Our take: Insurance software vendors have competed on their screens for twenty years. That is starting to change. When a producer can do the work from an assistant they already have open, the vendor’s interface matters less. What matters more is how complete the API is, how clean the data is, and how well the approval steps work. Carriers and agencies that still treat AI as a chatbot attached to a portal will fall behind the ones that make their system of record easy for agents to use. The risk grows along with the access. An agent that is allowed to act can cause a loss, which is why some security writers now suggest limiting how much each agent can do before a human has to approve it (Forbes).

What to check in your own stack:

  • Can an outside agent read and update the records your staff work in every day, or only a small subset?
  • Is every action an agent takes logged with the agent, the approving human and the before/after values?
  • Which actions need human approval, and who decides where that line sits?
  • Does your vendor’s roadmap include MCP or similar agent access, or only its own built-in assistant?

2. Insurtech funding fell 45%, and the details matter more than the headline

What happened: On November 5, Gallagher Re reported that global insurtech funding in Q3 2026 was $1.36 billion. That was 44.6% lower than Q2 but 34.9% higher than a year earlier. Funding rounds over $100 million totaled $660 million, down from $1.67 billion in Q2. There were 104 deals, compared with 107 the quarter before. AI-focused companies took 83.2% of all funding, and U.S. companies accounted for a record 64.4% of deals. (Business Insurance)

Our take: This isn’t a downturn. Deal count barely changed. The drop comes from Q2, which had an unusual cluster of very large rounds. The real story is how concentrated the money has become. Almost all of it goes to AI, and most of it goes to U.S. companies. Public insurtechs reported strong results the same week. Root earned $14.4 million in Q3, and Lemonade said it expects its first positive adjusted EBITDA quarter in Q4 (Insurance Journal). Together, these point to investors backing two kinds of companies: profitable full-stack insurers and AI infrastructure. Companies in between, such as digital agencies without their own underwriting or platforms built on older software, are having a harder time raising money.

What this means in practice:

  • Founders should be able to explain clearly what their AI changes in underwriting, claims or distribution economics.
  • Carriers choosing vendors should ask how much runway well-funded AI vendors have, and plan for consolidation among the rest.
  • Expect more acquisitions like Guidewire’s purchase of hyperexponential, as large platforms buy AI capabilities instead of building them.

3. Vertafore’s change-request agent: a useful test case for MGAs buying AI

What happened: On November 5, Vertafore released its Velocity Change Request Agent in Surefyre. The agent compares renewal and endorsement applications with existing policy data, summarizes what changed in plain language, and sends the result to an underwriter for approval before anything is saved. Vertafore says internal testing showed 70% less time spent reviewing and documenting changes, and 98% fewer errors. It is the sixth AI agent in Vertafore’s Velocity line for MGAs. (Insurance Edge)

Our take: Endorsements and change requests are good work to automate. There is a lot of it, it follows rules, and mistakes are expensive. That makes Vertafore’s choice a sensible one. But “internal testing” is a number from the vendor, not a result on your own book of business. MGAs are now being offered many agents, each with an impressive headline figure. Zywave’s new research puts conventional agencies at about $175,000 of revenue per employee, compared with close to $1 million at AI-native brokerages (Reinsurance News). That kind of gap will push buyers to move quickly. They should still check the claims first.

Questions for MGAs before buying:

  • Will the vendor run a pilot on your own submissions and report cycle time and error rate compared with your current baseline?
  • Where does the underwriter approve the work, and is that approval recorded in a way an auditor or regulator can follow?
  • Does the agent work across your systems, or only inside one vendor’s platform?
  • What happens to underwriter capacity, and do you plan to write more business or need fewer people?

Also from the week: Zywave’s “Broker of the Future” report estimates AI could create more than $100 billion a year in value across insurance distribution within five to seven years (Reinsurance News). Reliance Global Group, an insurtech holding company, is changing its Nasdaq ticker from EZRA to XGPU after buying GPU cloud assets, while its RELI Exchange insurance business continues (GlobeNewswire).