Industries / Finance and business services / Insurance / Insurance brokerage
Renewals
Renewal is the yearly round in which a broker reviews a policy before it ends, takes it back to insurers if needed, and tells the client what has changed in price and cover.
6
Milestones achieved, from 2002 to 2026.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
2002
A hard market, measured
A hard market is a period when insurers raise prices and narrow cover across the board. The Council of Insurance Agents + Brokers (CIAB), a US brokers' association, surveys its members each quarter on renewal prices. In January 2002 it reported rises through the last quarter of 2001, "following the Sept. 11 terrorist attacks", and "nearly half indicated rates for medium and large accounts grew 30 to 50 percent". A year later its president, Ken Crerar, said: "We are well into our second year of hard-market conditions."
Credit
CIAB; Ken Crerar
2015
Leave at any time
From 1 January 2015, under a French consumer law of 17 March 2014 called the loi Hamon, car, motorbike and home policies and add-on cover can be cancelled at any time after the first year, ending one month after the insurer receives notice, by the consumer group Que Choisir's account. French insurers must also state the deadline for cancelling on the renewal notice; if it arrives late, the policyholder has 20 days from its sending. In Germany, the Insurance Contract Act limits each automatic extension of a policy to one year.
Credit
French Parliament
2017
Last year's price shown
From 1 April 2017 the FCA required insurers and brokers to "disclose last year's premium at each renewal", to encourage shopping around, and to give an extra message to customers who had renewed four times in a row. A review in October 2017 found some firms sending "expiry" or "re-broking" notices that left the information out. In 2019 the FCA estimated "consumer savings of between £39 million and £330 million a year (with an average estimate of £185 million a year)", against costs of "around £4 million a year".
LessonShowing last year's price helped some customers, and in 2022 the FCA went further and banned charging renewing customers more than new ones.
Credit
FCA, policy statement PS16/21
2019
A freeze on dropping homes
In December 2019 California's insurance commissioner, Ricardo Lara, used a new law, SB 824, to bar insurers for one year from refusing to renew about 800,000 homes "in ZIP codes adjacent to recent wildfire disasters", the first time the department had used it. Non-renewals had risen by more than 10% the year before in seven counties. In the five years to March 2025 the California FAIR Plan, the state's insurer of last resort, "experienced 400% policy growth", and in 2025 its member insurers had to pay in a combined $1 billion.
LessonFreezing non-renewals for a year did not keep homeowners out of the last-resort plan.
Credit
Ricardo Lara, who wrote SB 824 as a state senator; California Department of Insurance
2024
Rates stop rising
Marsh's Global Insurance Market Index tracks commercial insurance prices at renewal each quarter. In the second quarter of 2024 rates were flat, "ending 26 consecutive quarter streak of increasing rates", "the first time in nearly seven years", by Marsh's count. By the third quarter of 2026 they had fallen for nine quarters in a row, by 5% in that quarter.
Credit
Marsh
2026
Notices that explain nothing
In August 2026 ASIC, Australia's financial regulator, reported (REP 838) on eight car insurance brands from five insurers, about 72% of the market. "None of the insurers reviewed explained the key factors that affected how premiums were calculated" in quote or renewal documents; 67% of customers renewed with the same insurer; and 31% of those who contacted their insurer had their premium cut with no change to cover. A review of 52 home insurance renewal notices by the consumer group CHOICE and the Financial Rights Legal Centre found that "0% of renewal documents provided an adequate reason for the premium increase".
LessonShowing last year's price is not the same as explaining the new one.
Credit
ASIC; CHOICE; Financial Rights Legal Centre
Watch for
Renewals without a submission
Applied Systems says AI from Cytora, which it bought in 2025, picks out renewals in its Applied Epic system that qualify and sends the risk data straight to insurers, which return renewal quotes, with "Travelers signing on as the first anchor carrier". It says quote turnaround fell "from weeks to hours" for some partners.
Achieved wheninsurers or agencies outside Applied publish the number of renewals quoted this way and the turnaround.
Claimed by
Applied Systems, July 2026
Renewal agents for agencies
Zywave released AI agents to "more than 35 early adopters" and said it would add agents for "new business and renewal quoting automation". It gives no accuracy figures.
Achieved whenan independent party measures the agents' error rates on real renewals.
Claimed by
Zywave, February 2026
Open
01
What changed at renewal
A public test set of 100 made-up pairs of policies, last year's and this year's, each with a declarations page, forms and endorsements. Together they carry 500 planted changes that matter (a lower limit, a higher deductible, a new exclusion, a narrower definition, a dropped endorsement) and 150 that do not (renumbering, new layout). Tools are scored on the changes they find, the decoys they wrongly flag, and how they rank severity.
The briefCredit · unclaimed02
Renewal notices checked
300 made-up renewal notices with planted defects, checked against published rules: the UK's (last year's premium, the shop-around message, the extra message after four renewals) and France's (the deadline for cancelling shown on the notice), plus ASIC's finding that notices should explain why the price changed. Defects include a missing prior premium, a prior premium buried on a later page, weakened shop-around wording and a notice labelled as an "expiry" notice. A checker reads each notice and lists its defects.
The briefCredit · unclaimed
Open 01 · the brief
What changed at renewal
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test set of 100 made-up pairs of policies, last year's and this year's, each with a declarations page, forms and endorsements. Together they carry 500 planted changes that matter (a lower limit, a higher deductible, a new exclusion, a narrower definition, a dropped endorsement) and 150 that do not (renumbering, new layout). Tools are scored on the changes they find, the decoys they wrongly flag, and how they rank severity.
Why it is still open
Policy-checking services sell this work but publish no tests: Exdion claims "unparalleled accuracy", and Patra "99%+ Platform accuracy with expert validation", with no method given. The nearest public benchmark, ContractScrub (August 2026), checks single contracts "hand-crafted by experienced lawyers", not two versions of one, and "only one model" reached 0.75 macro-average recall. Swiss Re Corporate Solutions' claims data for brokers' errors and omissions cover showed "renewal errors have increased while new business errors have declined".
Built with what exists
long-context AI models such as Claude, Gemini and ChatGPT (see the Note "Seven more flagged"); Python's difflib as a plain text-comparison baseline; a PDF generator such as ReportLab for the made-up policies.
First step
Write the list of change types, then build ten pairs by hand with 50 planted changes.
Done looks like
The set, answer key and scorer are public. Results for at least three AI models and the difflib baseline, with the share of planted changes found and false alarms on decoys for each. A second person checks the answer key on 20 pairs.
Open 02 · the brief
Renewal notices checked
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
300 made-up renewal notices with planted defects, checked against published rules: the UK's (last year's premium, the shop-around message, the extra message after four renewals) and France's (the deadline for cancelling shown on the notice), plus ASIC's finding that notices should explain why the price changed. Defects include a missing prior premium, a prior premium buried on a later page, weakened shop-around wording and a notice labelled as an "expiry" notice. A checker reads each notice and lists its defects.
Why it is still open
The reviews so far were done by hand: the FCA's of 2017, ASIC's of eight brands and CHOICE's of 52 notices. The closest product found, Nomad Data's, advertises checks of US non-renewal notices against state templates and publishes no results.
Built with what exists
document-reading AI models; the FCA's PS16/21 rules and the French rules as the checklist; a PDF generator for the notices.
First step
Write the checklist from PS16/21 and make 30 notices with 60 planted defects.
Done looks like
On 100 held-back notices, the checker finds at least 95% of the defects, with no more than 5% false flags. The notices, rules and scores are public.