The renewal decision is made months before the renewal date. It accumulates across the whole policy year: the onboarding call, the premium that failed in February, the claim that went quiet for two weeks, the coverage change nobody flagged. By the time the notice arrives, the policyholder is confirming a conclusion they already reached.

That makes renewal a test of whether support ran as a lifecycle function or a ticket queue. An agent that only exists when the customer initiates cannot pass it, because for most of the policy year the customer is not initiating anything.

The renewal window

Price is the reason policyholders give, but indifference is usually why they were shopping in the first place.

Acquiring a policyholder costs an insurer far more than keeping one, and every carrier knows it. The operational attention most insurers give to retention is still a fraction of what they put into acquisition. The renewal window, typically 60 to 90 days out, is the highest-leverage point in the term, and most of the industry treats it as a form letter and a premium notice.

The policyholder who feels heard, updated, and proactively served in the months before renewal does not shop around. The one who had to call three times to get a claims update, or whose payment failed silently last February, is already comparing quotes before the renewal notice arrives.

Service quality and churn

Accenture surveyed more than 6,700 policyholders across 25 countries about their recent claims experiences. Among the claimants who were not fully satisfied with how the claim was handled, 30% had already switched carriers in the previous two years and a further 47% were considering it (Poor claims experiences could put up to $170B of global insurance premiums at risk by 2027, Accenture, August 2022).

Claims are the only moment when a policyholder actually uses their product, which puts the relationship in play. When the claim goes badly, with slow updates, opaque decisions, and support staff who have no context on prior conversations, the policyholder learns something important about what they bought. They now have a concrete reason to leave.

Service quality between claims matters too. Most policyholders will not file a claim this year, so their whole experience of the carrier is made of smaller moments: a coverage review, a payment confirmation, a notification that arrives before they have to ask.

What actually moves retention

None of this work is complicated. It reaches the right policyholder at the right moment with the right information, without a person having to spot the trigger first. The hard part is that it has to run continuously, across the whole term, on whatever channel the policyholder picks up.

Renewal outreach carrying the policyholder's actual coverage details and any change to their premium does work a form letter cannot. A generic reminder gets ignored, while a message that says "your home policy renews in 45 days, and here is what changed on your premium and why" gets read as service.

Claims follow-up outreach closes one of the most common sources of post-claim churn. When an AI concierge monitors open claims and sends status updates without the customer having to ask, built on multi-agent coordination, the policyholder stops wondering whether anyone is working the file.

Payment failure recovery is a retention problem that rarely gets classified as one. Silent payment failures, where the customer does not know their policy is about to lapse, are responsible for a meaningful share of involuntary churn.

AI can identify a failed payment within hours and open the recovery conversation on whichever channel that policyholder uses, whether that is SMS, email, chat or voice. The concierge that texted them is the concierge that answers when they call back, holding the same context, so nobody re-explains a failed direct debit to a second queue. The alternative is a lapse notice, which damages trust even when the failure was not the customer's fault.

Coverage gap notifications address a different kind of failure. When a customer's circumstances change (a new vehicle, a home renovation, a teenage driver) and their coverage does not move with them, they are both underinsured and uninformed. A concierge that monitors policyholder data for coverage mismatches and opens the conversation reduces risk exposure and shows the policyholder that someone is paying attention.

Coverage is also where the determinism line has to sit. Spotting the mismatch is good work for AI, because it is pattern matching across policy data and life-stage signals. Stating what someone is covered for is regulated territory, where a wrong answer is a bad-faith exposure.

The split that works is AI to detect the gap and open the conversation, deterministic logic to state the actual coverage terms, and a licensed human for anything that amounts to advice. The same line runs through premium statements and lapse notices.

Anything a carrier would have to defend in a complaint file should come out of a deterministic path that produces the same words every time, pulled from the policy record rather than generated. In the coverage conversation, AI is a liability question before it is a retention one, and the answer has to be a mechanism the carrier can point at.

Complex inquiry handling (the coverage question with no simple answer, the claim dispute that needs explaining, the endorsement that involves several decisions) is where concierge-style AI support separates from a FAQ chatbot. A policyholder who asks a complicated question and gets a complete, accurate answer without being transferred twice remembers it, and earning that trust is how the relationship gets built. That experience becomes a reference point at renewal.

Proactive and reactive

Reactive support waits for the customer to contact the insurer. The insurer resolves the issue and the interaction ends. This model is operationally efficient and retention-blind.

Proactive support starts from a signal. The concierge watches claim age, payment status and engagement, opens the conversation when one fires, and carries the context forward when the policyholder answers on a different channel three days later.

The signals are specific: a claim open for 14 days without an update, a payment that failed three days ago, a policy nearing renewal with no engagement recorded, or a coverage profile that does not match the customer's known life stage. The outreach matches that policyholder's situation.

The difference is timing. Reactive support reaches a policyholder who is already frustrated, while proactive outreach reaches them before the frustration hardens into a decision. The cost of a well-timed message is a fraction of the cost of re-acquiring a churned policyholder.

Measuring retention impact

Renewal rate, the percentage of policyholders who renew at the end of their term, is the top-line number and a lagging one. By the time it moves, the work that moved it is months in the past, so it has to be read alongside leading indicators.

NPS tracked by interaction type shows where the relationship is being won or lost. A policyholder who had a claims interaction last quarter and gave a detractor score is at elevated churn risk. Outreach targeted at that cohort (following up, checking in, resolving outstanding issues) shows up in renewal rate a few quarters later.

Cancellation reason tracking, done rigorously, closes the feedback loop. When a carrier's non-renewal reasons cluster around silence after a claim rather than premium, the fix is operational and it is available now. AI outreach programs that address the specific reasons policyholders leave are the ones that change the renewal curve.

Where Lorikeet fits

Lorikeet runs concierge support on complex inquiries, trained once and deployed across chat, email, SMS and voice, with the same concierge holding context when a policyholder moves between them. In insurance that is live at SageSure, where the AI reads an email, updates the policy and replies to the customer without a human touching it, across a book of 850,000+ policyholders in catastrophe-exposed residential property.

"Not only can Lorikeet receive an email on my behalf, they could tell me what the email is, they could make updates to the policy, and they could reply to the customer without me ever needing it to cross the desk of one of my employees."

Pete Rizzo, AVP of Service and New Business Onboarding Optimization, SageSure

Proactive outbound is shipped. Carmoola, which operates in UK consumer credit regulated by the Financial Conduct Authority, resolves 90% of outbound conversations end-to-end with Lorikeet, and an A/B tested proactive outreach lifted one of their conversion metrics by 60%. The insurance version of that mechanism is the renewal notice that carries real coverage detail, the claims update nobody had to ask for, and the failed payment conversation that opens before the lapse notice does.

The retention case for AI is a sequencing problem. The work is reaching a policyholder at the point in the term when something changed, with information they can use.