The insurance follow-through platform
AI that finishes the work.
And proves it.
NewCover launches AI-run cohorts across your book — renewals, collections, claims documents — and follows every account to a documented outcome or an owned blocker. Every action shows its reason, its source data, and its approval trail.
Never a chatbotNever a black boxApproval-first by design
Brookline Dental Partners
Renewals · March cohort · due in 9 days
$48,000
Next action — send renewal reminder
Policy expires in 9 days. Two reminders unanswered. Premium up 12% — increase explanation attached from carrier rate sheet.
sources: policy:GL-4481 · rate-sheet:2026-03 · thread:re-renewal
⚠ guardrail: premium > $25K → approval required
Type APPROVE ↵✓ approval_granted · dana · reason recorded · audit AE-1204
01 — The follow-through gap
Nobody loses a renewal at the quote. They lose it in the follow-through.
Drafting tools write emails nobody sends. Dashboards report problems nobody owns. The work that actually moves money — the third reminder, the chased carrier, the collected payment, the signed affidavit — still dies in inboxes.
The renewal that expired quietly
Everyone assumed someone else was on it. The account found out from the carrier's lapse notice.
The payment that was promised
The customer said Friday. No one owned Friday. The receivable aged into a write-off conversation.
The claim stuck on one document
A loss affidavit sat unsigned for three weeks because chasing it belonged to nobody's job description.
Generic AI answers questions. NewCover exists for the other problem: making sure the work actually finishes — and being able to prove that it did.
02 — How it works
Follow-through, run as a system.
One operational spine under every module: Module → Cohort → Work item → Action → Outcome or blocked state → Attribution.
- 01
Pick a module
Fourteen operational modules, one contract: defined outcomes, defined blocked states, nothing open-ended. Most teams start with renewals — the deepest journey.
- 02
Launch a cohort
Import a segment of your book. Data-health checks flag missing outreach channels, do-not-contact conflicts, and stale records before anything can go wrong.
- 03
Simulate before anything sends
Every cohort dry-runs against your guardrails and workflow first. Launch requires an explicit human approval — outbound automation never starts itself.
- 04
AI works every item
Each proposed action carries its reason, its confidence, and the source data behind it. Trust levels decide what sends supervised and what waits for sign-off.
- 05
Your team judges — it doesn't triage
The Today queue surfaces only what needs a human decision: approvals, high-value risk, blocked data. Everything else keeps moving without you.
- 06
Outcome, or an owned blocker
Every work item ends documented: renewed and paid — or blocked, with a named owner and an aging clock. Attribution splits the credit: AI, human, shared, or outside.
03 — Modules
Fourteen modules. One contract.
Every module declares its outcomes and its blocked states up front. Renewals is the deepest journey; the rest of your operation is on the same rails.
Renewals
ActiveEvery policy to paid renewal, documented decline, or an owned blocker — before expiry.
Revenue
Collections
PilotOverdue premium followed to paid, promised, or escalated — with evidence.
Revenue
Retention & saves
PilotCancellation-risk accounts worked to saved or retained-with-change.
Revenue
Billing follow-up
On requestFailed payments, plan confirmations, and billing exceptions resolved.
Revenue
Quote follow-up
On requestOpen quotes chased to bound, received, or documented loss.
Coverage
Underwriting follow-through
PilotSubmission requirements from missing to quote-ready.
Coverage
Producer follow-up
On requestProducer-owned requests with due dates, reminders, and audit trail.
Coverage
Carrier follow-up
On requestUnanswered carrier asks tracked until response or escalation.
Coverage
Claims intake
On requestClaim notices turned into complete, routed claim packets.
Claims
Claims document chase
ActiveMissing claim documents collected; unresolved blockers routed to owners.
Claims
Policy servicing
On requestCertificates, endorsements, and service requests to completion.
Coverage
Reinsurance coordination
PlannedTreaty evidence and internal approvals, coordinated.
Control
Pulse
ActiveSupervisor proof layer — outcomes, blocked aging, attribution.
Control
Audit & compliance
ActiveEvery action's reason, approval, and state change — reviewable.
Control
Your workflow?
If it has an outcome and a blocker, it fits the contract. Pilot a new module with us.
Request a module →
04 — Proof
Every number defends itself.
Pulse is the supervisor layer: value protected, outcomes landed, blockers aging, attribution split. Each metric links back to the cohorts, work items, and audit events that produced it. If a number can't show its evidence, NewCover doesn't show the number.
Pulse · platform metric
Value protected or recovered
$1,605,600
attribution · AI 46% · shared 27% · human 18% · outside 9%
✓ renewed_paidEvergreen Studio · $61,200 · AE-1198
✓ paidUnion Street Books · $18,400 · AE-1201
⚠ blocked · waiting_on_carrierFoxtail Grocery Co. · owner: marcus · aging 4d
Demo environment shown. Your Pulse is built from your cohorts — and only from evidence.
05 — Trust
Built to be audited, not believed.
Insurance is a regulated business. NewCover treats every AI action as something a compliance team will one day read — because they will.
Approval-first trust levels
Draft-only, approve-before-send, supervised — you set how much authority the AI has, per module. It never grants itself more.
Every action cites its sources
AI actions must reference the records they're based on. An action with no source data is invalid by contract — it cannot run.
Audit written with the change
State changes and their audit events commit in the same database transaction. There is no window where something happened but wasn't recorded.
Typed confirmations
Assigning, approving, deferring, recording an outcome — every state change requires an explicit typed confirmation and a recorded reason.
No double-sends
Commands are idempotent. A retried request replays the original result instead of acting twice — even under concurrency.
Isolated, signed, expiring
Tenant-scoped data end to end. Customers act through signed, expiring secure links — no logins, no exposed portals.
What NewCover will never do
- ✕Send outbound without passing your guardrails and approval policy
- ✕Show a metric that can't link back to its evidence
- ✕Act on an account with no source data behind the action
- ✕Replace your AMS, your CRM, or your judgment
- ✕Pretend a chatbot is an operations team
06 — FAQ
Direct questions, direct answers.
Is NewCover a chatbot?
No. NewCover is a follow-through system. Conversations exist inside it as evidence attached to work items — not as the product. The unit of work is an account moved to an outcome, not a message answered.
Does it replace our AMS or CRM?
No. NewCover runs alongside your systems of record and takes on the operational follow-through they don't do: chasing, reminding, collecting, documenting, and proving what happened.
What happens when the AI isn't sure?
Low confidence is a blocked state, not a guess. The item routes to a human with the AI's reasoning and source data attached, and nothing sends until someone with authority approves.
How do you know what the AI actually contributed?
Every final outcome carries an attribution split — AI, human, shared, or outside NewCover entirely — recorded with the outcome and backed by the item's full audit trail. If a renewal was going to renew anyway, the attribution says so.
What does a pilot look like?
One renewal cohort from your real book. Import, data-health review, simulation, launch with your approval rules — then watch every item land on an outcome or an owned blocker. You keep the evidence either way.
How is our data protected?
Tenant isolation at the database layer, signed session auth on every API read, signed expiring links for customer actions, and an append-only operational audit trail. Security posture is documented and inspectable during the pilot.
Pilot on one cohort. Keep the evidence.
Put your book on follow-through.
Start with one renewal cohort from your real book. Watch every account land on an outcome or an owned blocker — and see exactly who earned what.
pilot@newcover.ai