Lending Agent Builder Author it like a playbook. Version it like code.

• Lenders who want to define their own lending agent — steps, SLAs, document requirements, escalation — without touching TE's code.
• Compliance and ops, who need an audit trail: who approved what, and when.
• Borrowers, whose self-serve experience is gated by a plain-language policy the lender's admin writes, not hardcoded per channel.

• A structured playbook spec: steps, each with an execution mode (autonomous, act then notify, LO approves), SLAs, and per-branch or per-LO tier overrides.
• Version lifecycle draft → shadow → pending approval → live → retired, with diff-based history and one combined approval gate per version.
• Shadow mode runs a draft playbook against live traffic without taking real actions, so it can be compared to the live version before it goes live.
• What the borrower can see or do while the playbook runs is gated separately, by the Loan Persona Builder — not hardcoded into the playbook steps.
• Pricing and fees are agent-callable, backed by either the external PPE or Agentic PPE (the Context layer's PPE integration for agents) — same shape either way, so the playbook doesn't need to know which one answered.

first-time buyer, gift funds — the first reference flow to build and shadow-test against

1. A lead is created for a first-time buyer using gift funds.
2. A decision model assigns the lending agent (playbook) for that scenario.
3. The orchestration engine deploys the scenario, SMS and sales assistant.
4. The application starts via the POS integration; the agent captures details from a voice and SMS call.
5. POS is updated with details from the lender's own consumer UI.
6. The application is submitted; the LOS integration updates the orchestration engine.
7. An SMS notification goes to the LO.

Once shadow mode ships, this flow is both documentation (a sequence diagram) and a shadow-mode regression test run against every new playbook version — not just a diagram that goes stale.