AiNOS Keep

Persistency intelligence. Keep it in force.

Why a book lapses, named policy by policy, and the interventions worth a call ranked first.

Deterioration is decomposed into drivers with evidence for each, and interventions are ranked by what a call can change. The retention team decides.

Why is this book deteriorating, and what should we do first?

Keep starts with persistency and retention, in Policy Servicing & In-force.

What the policy has to survive: kept long enough to pay back what it cost to sell.

Thirteenth-month and twenty-fifth-month persistency arrive as a report. By then the cohort is gone.
The causes are mixed together: product fit, agent behaviour, commission timing, customer events and service failures all show up as one lapse rate.
Retention teams work a list sorted by premium, not by the chance that an intervention changes the outcome.

Every in-force policy carries its own history: who sold it, under what scheme, what changed since, what the customer asked for.
Deterioration is decomposed into named drivers with the evidence for each, so a product issue and an agent issue are not treated the same way.
Interventions are ranked by expected effect and cost, and the retention team records what it did and what happened, which trains the next ranking.
The same view answers the compliance question when a lapse pattern points to how the policy was sold.

None of this works on raw legacy tables. These are the parts of AiNOS One it draws on.

Policy, agent and customer histories resolved into one timeline from administration, commission and service systems.
Commission scheme and product rules compiled so their effect on lapse can be tested.
Outcome records kept so each intervention becomes evidence for the next.

Prepare
The AI explains deterioration and proposes a ranked list. The retention team chooses the interventions.
Govern
Early-warning runs over the whole book continuously. Interventions follow approved playbooks with recorded outcomes, and exceptions go to a person.
Connect
Persistency findings change commission design, product rules and conduct scoring, because the loops share one picture of the policy.

What Keep learns is handed to the other two, and the rest of policy servicing & in-force is built on the same cases.

To AiNOS Bind: A product that lapses at month 13 for everyone changes what Bind checks before the next one is written.
To AiNOS Guard: A lapse pattern that follows one agent is handed to Guard as a sales-conduct case.

Owned by policy administration, customer service, retention and conservation.

Workflow	Today	AI-native
Customer inquiries	Contact centre answers depend on which agent picks up and which screen they have open.	Answers are grounded in that customer's policy, its history and the rules that apply, with the source shown to the agent.
Policy transactions and notices	Changes, loans, surrenders and reinstatements are processed against rules the processor has to remember, including legally required notices.	Each transaction is checked against the contract and the notice obligations in force that day, and the notice is produced with the transaction.
Collection and lapse	Arrears are worked from a list, in order, after the grace period has started.	Lapse risk is scored as arrears begin, with the likely reason, so the outreach is specific and early.
Customer lifetime value	Cross-sell is driven by campaigns rather than by what changed in the customer's life.	Changes in a customer's needs are detected from their own events, and any offer is checked for suitability before it is made.