Is the ACORD form dead, and is legacy still winning? Branch, Ryan Specialty, Tokio Marine, AmTrust and Upstage on Day 2 of Insurtech Insights USA 2026

Should the industry kill the ACORD form?

The panel title was a provocation, and the panel said so. Moderator Mang-Git Ng of Anvil, which turns PDFs into APIs, asked whether the industry has a data problem or a PDF problem.

Chad Nitschke of MGT Insurance took the antagonist’s chair. At the carrier level, he said, MGT almost never sees forms. It pulls from eleven data sources and about 300 data fields beyond what an ACORD form holds. Its underwriting agent, Athena, sums up that data in about two minutes and hands the underwriter a recommendation and a price. “We always have a human in the loop,” he said.

Cameron Nordholm of Covered pushed back. Covered works with more than 70 carriers and tens of thousands of quotes a month. To him, ACORD is the semantic layer of the market, the shared definition of how the industry talks to itself. If agents are going to talk to carrier systems, he said, the industry will lean on those definitions more, not less. His advice: don’t throw the baby out with the bathwater.

Dan Clark of Pathpoint found the middle ground. ACORD does not have to be a PDF. It can be XML or JSON. What matters is the schema. Jason Keck of Broker Buddha agreed the data model has real value but said the output format has to change. He also gave the room a warning about AI.

“AI agents are incredibly confident in their output, but also they’re often wrong.”

Jason Keck, Co-Founder and CEO, Broker Buddha

Asked who wins from here, the panel picked nimble companies with deep domain knowledge over both large software vendors and horizontal AI firms. Ng’s summary was blunt: when an incumbent says it is AI native, don’t believe it.

How did Branch launch in 40 states so fast?

Joe Emison, co-founder and CTO of Branch, told the story of a push that ran from 2019 into 2022. Branch got new admitted programs approved in 40 states in under 40 months. Branch has since begun managing older farm mutuals and bringing their technology up to date, and he said many carriers ask how it was done.

His answer had two parts. First, people. In older companies, one person’s job slowly becomes the job of fifteen or twenty people, and launching anything takes a lot of coordination. Second, technology. Most core systems, he said, were built by vendors copying the habits of a first big customer. He told a family story about cutting the Thanksgiving turkey in half for years because grandma’s oven was once too small.

Branch’s fix was what he called a “techno-social” platform. Small teams own a whole job from end to end. Branch’s forms team drafts, files, loads and tests its own forms. Its rating team of two handles filings, loads rates and certifies they are right. Branch bought off-the-shelf tools for messaging and customer service. Something like 95% of its code, he said, is front end and glue. It did all this with fewer than ten developers, most trained in its own boot camp in Columbus.

He also pushed back on a claim from the day before that every company will soon build its own policy admin system with AI. Code is the easy part, he said. A real platform takes about 2,500 cycles of building, putting it in front of users and watching what happens. Branch now releases every night behind a full automated test suite.

“Every time you put your product in front of users, you get punched in the face because they never use it the way you think they will use it.”

Joe Emison, Co-Founder and CTO, Branch

Why is legacy so hard to leave?

Tony Lew of InsurTech NY moderated “Is Legacy Still Winning?” with four panelists from carriers, advisory and payments.

Tony Skipper, former CEO of Allianz Technology of America and former CIO of AIG Japan, listed the costs of legacy. Security comes first. Then come blocked products, hard integration, scarce talent and missing documents. His advice for making the case to replace it: “Take off your technology hat and put on your business hat.”

Raj Kalahasthi of Catalyx Advisory said the real block is the operating model and the courage to change it. Early in his CIO career, a payments upgrade built from the inside out failed and had to be restarted from the customer’s view. Later, at a fast-growing broker, he said cultural alignment mattered more than the modern stack.

Reuven Shnaps of AmTrust said data is everything. Carriers sit on years of data, but in silos, with different meanings for words like premium and profit. His fix: break the silos, agree on terms, and start small and modular instead of trying to boil the ocean.

Josh Peterson of ePayPolicy brought the good news. One company moving old desktop apps to React and APIs has cut two years off a five-year plan. Another finished a 12-month .NET migration in eight. AI, he said, gets the answer key: the old code, the documents and the tests.

How far ahead of governance is AI?

Richard Hartley of Cytora moderated the AI Maturity Curve session. Bob Pick of Tokio Marine North America said his team took its first generative AI tool, a document comparison tool, live by the end of 2023. His caution: AI capability is about a year ahead of the industry’s ability to govern and control it.

Pick said one group carrier runs six policy admin systems and has put Salesforce over all of them as one skin for underwriters. Rating, forms and AI tools live in point products around it. “The core gets thinner and thinner,” he said. He added that teams must get used to shorter cycles: “You might live with your AI solution for twenty five months, or five months, and that’s okay.”

Murad Lodhi of IFG Companies said the business has to drive the system. If the middle of the business is not sold, don’t start. On one agentic project, he said, quotes doubled and submissions tripled, and he wants the market moving toward same-day quoting. IFG is about 15 months into a three-year roadmap, swapping the “underwriting workbench” for an “underwriting dashboard” and retiring RPA bots from 2017.

First mover or smart mover?

Kristoffer Lundberg of Insurtech Insights moderated this session with Kasey Roh of Upstage. Roh noted that Facebook changed its motto in 2014 from “move fast and break things” to “move fast with stable infrastructure.” Lundberg pointed to Bent Flyvbjerg’s research on big projects: the ones that work think slowly and act fast.

Roh said many carriers jump to ten agents in ten lines and never reach production. The better path is to pick one painful, high-volume workflow, set up a sandbox with real data and agree on what success looks like. For reading documents, Roh said, a small model running on two GPUs can do the job without frontier-model token bills. Agents should follow the carrier’s own step-by-step rules and stop when something is missing.

On ROI, Roh said only two large carriers gave specific, measurable AI results on first-quarter 2026 earnings calls. Humans stay in the loop, Roh said, because their feedback is what makes the model better. “You cannot outsource your thinking.”

What does Ryan Specialty’s new tech stack look like?

In a fireside chat with Mladen Subasic of Bound AI, Bradley Bodell, CIO of Ryan Specialty, said the core should become headless, opened up through APIs and kept in the back office. Ryan Specialty’s Empower program aims to shrink its tech stack after many acquisitions and help “brokers be brokers” and underwriters be underwriters.

Underwriters and brokers now bring tools they built over a weekend. So Ryan Specialty is building an AI sandbox with approved tools and data. Good ideas get paired with a forward-deployed engineer, then go through testing, security checks and review by legal, compliance and procurement. Bodell wants the speed of a sandbox with production discipline. “I do think we’re all going to become technologists,” he said.

Topics