BRENNEN

Industries / Finance and business services / Insurance / Insurance brokerage

Quoting

Quoting is gathering a client's details and getting prices and terms back from insurers.

5

Milestones achieved, from 1983 to 2026.

2

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1983

    A network to many insurers

    IVANS was set up as a shared electronic network between independent agents and insurers, so that policy data could pass between their systems. Rough Notes dates it to 1983, as a sister organization of ACORD, the insurance data standards body; Carrier Management says it was founded in 1984 by a consortium of 24 insurers. By 2013 about 350 insurers used its data download service.

    roughnotes.comcarriermanagement.com

    Credit

    IVANS and the insurers that founded it

  2. 1984

    Many prices, one entry

    A comparative rater lets an agent enter a client's details once and see prices from several insurers side by side. Rough Notes dates the CAIR II "Personal Lines Comparative Rating System for PCs" to 1984. Insurance Technologies Corp. says it launched its rater, TurboRater, in Texas in 1985. Neither source claims to be first.

    roughnotes.comturborater.com

    Credit

    CAIR Systems; Insurance Technologies Corp

  3. 2000

    Portals bring back retyping

    ACORD's AL3 standard let agents key in a client once and send the data to every insurer; by 2000 it had "more than 12,000 upload and 40,000 download implementations", by Insurance Journal's count. Then insurers "reverted back to proprietary applications", their own websites, where agents had to "rekey the data that was already in their databases". ACORD and ACT, a council of agents working on technology, had begun a standard in XML, a format for structured data, to fix it in 1999.

    LessonWhen each insurer builds its own website, the retyping falls on the broker; in 2024, 55% of agents quoted commercial cover through insurers' portals but only 32% preferred to, by IVANS's count.

    insurancejournal.comcoverpages.orgivans.com

    Credit

    ACORD; ACT

  4. 2010

    A shared portal stalls

    In 2004 six large UK insurers, Allianz, Aviva, AXA, Groupama, RSA and Zurich, formed imarket, a free website where brokers could get quotes from several of them at once. Six years later Insurance Times reported that it "has failed to attract the numbers that insurers envisaged", and its managing director, Martin McLachlan, said: "We expected higher numbers by now, say 50% higher."

    LessonInsurers can build a shared trading site, but brokers decide whether it gets used.

    insurancetimes.co.ukinsurancetimes.co.uk

    Credit

    Polaris, which ran it, and the six insurers

  5. 2026

    A price without terms

    In September 2026 Insurify, a US insurance comparison site, blocked Meta's Muse AI agent from its platform because "Muse presents only price and omits other details such as limits, deductibles, discounts, eligibility, and disclosures". Its co-chief executive, Giorgos Zacharia, said: "A quote without its context is not a fair comparison."

    LessonA quote is a price together with its terms, and a tool that passes on only the price can mislead the buyer.

    claimsjournal.com

    Credit

    Insurify

Watch for

  1. Forms that fill themselves

    Applied Systems says Epic Conductor, an AI feature of its Applied Epic agency system, makes "ACORD applications fill themselves in from submission documents", cuts submission handling time by 60%, returns renewal quotes "in 3 hours instead of 3+ days", and catches "a wrong limit or missing endorsement" when comparing quotes.

    Achieved whenan agency association, auditor or researcher publishes times and error rates measured on live submissions.

    insurance-edge.net

    Claimed by

    Applied Systems, October 2026

  2. Submissions built by agents

    Vertafore says its AI agents "build carrier-ready submissions, compare quotes, prepare proposals and check issued policies". The only figures it gives, from early testing, are for other tasks, such as document review time cut by "as much as 90%".

    Achieved whenan independent party measures the accuracy of the submissions and quote comparisons.

    fintech.global

    Claimed by

    Vertafore, September 2026

Open

  1. 01

    Quote portals for browser agents

    Rebuild the question flows of five insurers' public small-business quote pages as local copies: the same questions, branches and error messages, without their code or branding. Write 100 made-up businesses and score browser agents, AI that operates a web browser, on filling every copy correctly from one submission. Each copy has a marked stop line before any contact or payment page.

    The brief
    Credit · unclaimed
  2. 02

    Quote differences, planted

    A public set of 120 made-up risks, each with three made-up quotes, 360 in all, carrying about 600 planted differences in limits, deductibles, sublimits, exclusions, endorsements, dates and the name of the insured, plus harmless differences in layout as decoys. A checker lines the quotes up and lists each difference that matters.

    The brief
    Credit · unclaimed

Open 01 · the brief

Quote portals for browser agents

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

Rebuild the question flows of five insurers' public small-business quote pages as local copies: the same questions, branches and error messages, without their code or branding. Write 100 made-up businesses and score browser agents, AI that operates a web browser, on filling every copy correctly from one submission. Each copy has a marked stop line before any contact or payment page.

Why it is still open

FormFactory, a research benchmark of web form filling published in June 2025, covers general forms, not insurance quotes, and found that "no model surpasses 5% accuracy". Gaya, which sells portal-filling tools to agencies, wrote that web agents "are evaluating themselves against a simplistic benchmark (WebVoyager) that is far from reflecting the complexities human insurance agents and CSRs deal with."

arxiv.orgcatalyit.com

Built with what exists

Playwright to build and serve the local copies; Browser Use, an open-source browser agent (see the Note "Deleting 80% to start over"); computer-use models from Anthropic, OpenAI and Google. Local copies mean no made-up data is sent to real insurers.

  1. First step

    Record the questions, branches and error messages of five public quote flows by hand, and write 20 made-up businesses.

  2. Done looks like

    500 runs, 100 businesses on five portals. The best agent fills at least 99% of fields correctly and completes at least 95% of runs, and no run crosses the stop line. A person's time for the same 100 submissions is published alongside.

Open 02 · the brief

Quote differences, planted

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public set of 120 made-up risks, each with three made-up quotes, 360 in all, carrying about 600 planted differences in limits, deductibles, sublimits, exclusions, endorsements, dates and the name of the insured, plus harmless differences in layout as decoys. A checker lines the quotes up and lists each difference that matters.

Why it is still open

Applied and Vertafore sell quote comparison but publish no accuracy figures. The closest public test, Coverage Cat's AI leaderboard, says its "scenarios and answer keys were AI-authored and have not had independent insurance-expert review". LongListBench and SnorkelUnderwrite test extraction and underwriting, not differences between competing quotes.

coveragecat.cominsurance-edge.net

Built with what exists

document-reading AI models such as Claude and Gemini (see the Note "The jobs agents repeat"); a PDF generator such as ReportLab to make the quotes; Label Studio for the answer key. The quotes are written from scratch, not copied from ACORD's licensed forms.

  1. First step

    Write ten risks with 30 quotes and 50 planted differences, and time two people finding them.

  2. Done looks like

    The checker finds at least 95% of the planted differences, and at least 90% of what it flags is real. Five people, such as insurance students, are given the same set and time, and their scores are published alongside. A second person checks the answer key.