Bitvyas
All articles

Insurance Renewals with AI: What It Works Through, What Your Brokers Still Decide

By Akshit Agrawal, Co-founder, Bitvyas · Published 16 October 2026 · Facts as at 16 October 2026

At Bitvyas, we build insurance renewal automation as a workflow where AI works the renewal calendar 90, 60 and 30 days out, collects updated details from each client, assembles the comparison pack and issues routine certificates of insurance from templates. A licensed person keeps the advice, any change in risk and the renewal conversation.

Many agencies and brokers still run renewals from a calendar, a spreadsheet and an inbox. It works until a busy month, when a policy lapses quietly or a certificate request interrupts an afternoon. This post maps the process in the same order as our hub article, What to automate with AI first: the process, where AI takes it over, and what stays with a person. It covers a US independent agency and a UK broker.

What are the key facts about insurance renewal automation?

FactDetailSource
Certificate form (US)ACORD 25 (2025/12 edition) is the standard certificate of liability insurance; it states it "confers no rights upon the certificate holder" and does not amend, extend or alter coverageACORD 25 (2025/12), as published by NY DFS
UK renewal informationFor consumer general insurance contracts of 10 months or more (not group policies), the firm must give the renewal premium, a comparison with the original premium, a prompt to check cover and compare prices, and whether the policy renews automaticallyFCA Handbook, ICOBS 6.5.1R
UK renewal pricing ruleConsumer home and motor renewals must not be priced above the equivalent new business price; applies from 1 January 2022; commercial policies are excludedFCA PS21/5
What AI can do hereSchedule, collect, assemble, draft, route and trackBitvyas workflow design
What AI must not doGive advice on cover, accept a change in risk, or bind a policyBitvyas workflow design
Bitvyas client resultsNot published; we do not release client figures without permissionn/a

All as of 16 October 2026.

What does a renewal look like when it is done by hand?

The renewal date sits in an agency management system, a spreadsheet or a calendar. Someone checks it, emails the client for updated details, waits, chases, then prepares a comparison and a conversation. Certificate requests arrive in the middle of this, usually marked urgent.

Three weak spots repeat. The first reminder goes out late because the person who owns the calendar is busy. The client's updated details arrive in pieces over several emails. And a certificate request takes a producer or account manager away from the renewals that matter. These are volume and timing problems, not skill problems, which is why this process passes the four-question test in our hub post.

Which part does AI take over?

AI takes the clock, the collection and the first draft. Specifically:

  1. Work the calendar. Each policy is checked against its renewal date. At 90, 60 and 30 days out, the client gets the right message in the right channel, and the account manager sees who has replied.
  2. Collect updated details. The workflow sends a short form or a pre-filled summary of what you hold, and reads the replies, including emailed documents such as fleet lists or payroll figures. Gaps are listed, not guessed.
  3. Assemble the renewal pack. Expiring terms, the client's updates and the quotes received are put side by side in one summary, with the source of each figure.
  4. Issue routine certificates. A request that matches a client, a holder and an existing policy produces a certificate from your template as a draft. In the US that is typically an ACORD 25; the draft is checked against the policy before it goes out.
  5. Track and escalate. Anything unanswered, unusual or close to the date moves to a named person with the history attached.

The workflow works inside the system you already use, whether that is a US agency management system or a UK broker platform, and reads and drafts rather than replacing it.

What stays with a person?

A licensed person keeps four things. They give advice on cover: what to keep, drop or increase. They accept or decline any change in risk, such as a new location, new drivers or a change in turnover. They hold the renewal conversation with the client. And they approve any certificate that goes beyond the template, for example a request for wording the policy does not support.

A certificate is a record of cover, not a way to change it. The ACORD 25 says so on its face, which is why the check against the policy matters more than the speed of issue. The rule we build to: AI proposes with the evidence beside it, and a named person approves anything that binds, changes or promises cover.

How do UK rules change the workflow?

They change the content of the messages and the pricing checks, not who is accountable. For consumer renewals, the FCA requires the renewal premium, a comparison with the previous premium, a prompt to check cover and whether the policy renews automatically (ICOBS 6.5.1R). Home and motor renewals also sit under the equivalent-new-business-price rule in PS21/5.

A workflow can generate the required information in the notice and flag any renewal where the price looks higher than the equivalent new price. The firm still owns the pricing decision and the compliance sign-off. Commercial renewals are outside the consumer pricing rule, so a broker handling SME or fleet business uses a lighter notice flow. This is general information, not regulatory advice; check your own permissions and the current Handbook.

How does a month look in practice?

Take an illustrative example, with assumed numbers and not a client result. An agency has 300 policies renewing across a quarter. The workflow has already sent the 90-day messages and read the replies. The account manager opens one list for the week: policies with complete details, policies waiting on the client, and policies with a change in risk marked for a conversation.

A complete one gets its pack and goes to the producer. A waiting one gets a second message. A change in risk moves to a call. Meanwhile, five certificate requests arrive; four match a template and are drafted for a quick check, and one asks for non-standard wording and goes to a person. Nobody opened the calendar to find out what was due.

What can go wrong, and how is it controlled?

  • Wrong data read from a client document. Controls: show the source line next to every extracted value; send low-confidence values to a person.
  • A certificate that promises more than the policy gives. Controls: check wording against the policy before drafting; never auto-send non-template wording.
  • Reminders that annoy a client. Controls: set cadence and channel per client, and stop the sequence when the client replies.
  • Missing a lapse. Controls: anything unanswered at 30 days moves to a named person, and the workflow reports what it did not finish.

Is this worth building?

It depends on renewals per year, certificate volume and what a missed renewal costs you. Use the formula in What AI automation costs, and how to work out payback: build cost divided by the monthly saving after run and review costs. If you handle a few dozen policies, a calendar is enough. If renewals and certificates take a large share of your account managers' week, it is worth mapping. Our use-case page is AI handles policy renewals.

Related

Sources

  • ACORD 25 (2025/12) Certificate of Liability Insurance, as published by the New York DFS — https://www.dfs.ny.gov/system/files/documents/2025/12/acord-25-2025-12-liability.pdf
  • FCA Handbook, ICOBS 6.5 Renewals — https://handbook.fca.org.uk/handbook/icobs6/icobs6s8
  • FCA Policy Statement PS21/5, General insurance pricing practices, May 2021 — https://www.fca.org.uk/publication/policy/ps21-5.pdf

Figures and rules as of 16 October 2026. The month example uses assumed numbers for illustration. General information, not insurance, legal or regulatory advice.

Still doing this by hand?

Bitvyas builds AI that takes this kind of work off your team. It reads the documents, does the matching and drafts the next step, inside the tools you already use. A person still approves anything that matters.

Tell us what your team repeats