Platform

How Swivel works.

Swivel reads the systems you already run, builds a customer digital twin for every account, and gives specialized retention engines the context to act.

Talk to us Learn more about the work trial

1. Data in

Start with the systems you already have.

Swivel reads the places where retention signals already live: CRM, collection systems, your warehouse, and external context.

CRM, collection systems, data warehouse, and external data feeding Swivel.

CRM: Accounts, owners, plan, support history, and team notes.
Collection systems: Invoices, failed charges, dunning state, and renewal risk.
Data warehouse: Usage, logins, and engagement trends.
External data: Outside signals that sharpen the customer picture.

2. Customer model

A digital twin of every customer.

Specialist agents clean and join the signals, then keep a living account model that the rest of the system can read.

Data Cleaning Agent
Makes the incoming signals usable.
Aggregation Agent
Joins billing, product, support, and CRM context.
Digital Twin Modeler
Keeps the customer digital twin current.
Agent Orchestrator
Routes the work to the right engine and the right human.

Each twin keeps the context that matters

Behavior

Billing

Engagement

History

3. Retention engines

Specialists act on the twins.

Each engine owns one retention job. They run in parallel, using the same customer context instead of asking one general agent to do everything.

Retention engines around an intelligence hub: payment recovery, churn risk, cancellation saves, win-back, auto renewal, and onboarding.

Churn Risk Radar: Finds accounts pulling away before they cancel.
Win-Back Engine: Reaches lapsed customers with the right timing and offer.
Payment Recovery Engine: Stops billing problems from becoming churn.
Cancellation Save Engine: Matches the cancel moment to the right save play.
Onboarding Autopilot: Gets new customers to first value.
Auto Renewal Engine: Clears renewal friction before the term lapses.

Why this shape

A team is easier to understand and control.

The architecture stays legible: one shared customer model, specialist engines, and human approval before anything reaches a customer.

01
Specialization
Each agent has a narrow job, so the behavior is easier to review and improve.
02
Parallel coverage
The engines work every customer twin at the same time instead of waiting in a queue.
03
Human-approved handoffs
Agents prepare the save. Your team approves the customer-facing action.

Governance

Every customer-facing action stops at Sign-off.

Agents can prepare the work, but they do not send customer-facing actions without approval.

Security & governance

Sign-off: Approve, edit, or reject the action before it reaches a customer.
Reasoning trace: See the signals the agent used and why it suggested the move.
Audit trail: Every proposed and approved action is logged for review.

Tools we read from