Retention
Echo
The save happens before the lapse notice, not after
Echo watches the signals that precede a lapse — payment behavior, service contacts, life events, engagement decay — and runs the intervention while it still changes the outcome.
Interactive demo
Watch Echo work a real case.
Step through it one action at a time. Nothing here is a mock-up of a chat window — these are the actual stages of a run, including the checks that happen before anything visible occurs.
A 6-year whole life policy 58 days from a probable lapse
A policy in force since 2020, always paid on time, no service contacts. Nothing on the standard dashboard flags it. Echo scores it in the top 2% of lapse risk for the coming quarter.
Full transcript & decision trace
The part most vendors don't show you.
The verbatim conversation plus the reasoning trace behind it: rules fired, thresholds crossed, costs incurred, and every decision the agent deliberately refused to make.
What it's for
The problems Echo exists to solve.
Retention starts at the lapse notice
By the time a policy is in grace, the decision was made weeks earlier. The save campaign is arguing with a settled mind.
Save offers go to people who weren't leaving
Untargeted retention spend concentrates on the easiest-to-reach, not the most-at-risk — and looks successful because they stayed anyway.
Cross-sell is a list, not a signal
Campaigns run on product logic instead of the household events that actually create demand.
Capabilities
What it actually does.
Lapse prediction
Ranks the in-force book by 60-day lapse probability using payment, service, engagement, and life-event signals.
Intervention selection
Chooses the treatment — billing change, coverage review, producer call — matched to the reason, not a single generic offer.
Holdout measurement
A permanent randomized holdout means persistency lift is measured, not asserted.
Renewal orchestration
Drives renewal sequences across email, SMS, and Orion voice contact with frequency caps.
Cross-sell qualification
Surfaces households whose signals indicate genuine new demand and routes them to producers.
- Policy admin
- Billing / payment processor
- CRM
- Marketing automation
- Orion (voice)
- A permanent 5% holdout is never contacted — the only way lift stays honest.
- Frequency caps across all channels; Echo cannot out-contact your service team.
- No retention offer that a producer could not have made; pricing concessions require human approval.
- Vulnerability signals suppress automated campaigns and route to a human.
Human handoff: High-value or complex saves route to the servicing producer with the predicted reason and the recommended treatment.
Questions
Straight answers.
Why keep a permanent holdout?
Because without it, every retained policy looks like a save. The holdout is the only honest measure of whether Echo does anything at all.
Can Echo offer discounts?
Not on its own. Pricing concessions require human approval — most saves don't need one anyway.
How far ahead is the prediction useful?
Sixty days is the practical horizon where the intervention still changes the outcome. Earlier is noisier; later is arguing with a decision already made.
Keep going
The rest of the roster.
Orion
The always-on voice agent for outbound and inbound calls.
Quark
Sub-second multi-product quoting with rider optimization.
Atlas
Reads the file, runs the risk, books the clean cases.
Kairos
Pays the clean claims fast, escalates the rest in under 90 seconds.
Sentry
Watches every agent action and builds the audit trail.
Vesta
Closes tier-one service cases instead of deflecting them.
Meridian
Runs the escalation while the human makes the judgment call.
Bedrock
Turns the submittal and RFI log into a schedule early-warning system.
Keystone
Assembles change order entitlement and turns inspections into proposals same day.
Anchor
Calls every foundation inquiry in seconds and books the inspection.
Ridgeline
Plans the inspection and production boards together, and re-plans them when the day breaks.
Slate
Turns inspection notes, photos, and measurements into a priced proposal before the inspector leaves.
Run Echo against your own workflow.
Bring us one process and a number attached to it. We'll tell you whether an agent moves it.