Same queues
Agents pull work from the same smart queues as your team — routed by the same priorities and SLAs, visible in the same command center.
AI Agents
AI that handles servicing work within the limits you set
LendEasy agents work in the same queues and audit trail as your team, with their own scoped access and human approval where policy requires it. Each action is checked before execution and recorded afterwards.
Collections call — AI Agent
Case #C-2417 · first-party · early delinquency
Compliance gate — cleared before the dial
- 7-day call budget · 2 calls remaining
- Borrower local time · 2:14 PM — inside window
- Consent · on file — voice, cell
Live transcript
“…I can set that up. Your past-due amount today is $215.00 — would two payments, starting Friday, work for you?”
$215.00 · from the lending core
Recorded commitment
Promise to pay — 2 × $107.50, starting Fri
Within configured authority: 30-day horizon · minimum amount · installments
Pre-dial checks · core-supplied figures · promise recorded
AI in the servicing workflow
Give each AI agent its own access and task limits
Each agent has its own identity, scoped permissions, approved task types, and human review where policy requires it.
Separately scoped access
Each agent has its own identity and explicit permissions. Its access can be narrower than a person's role and is audited separately.
Same audit trail
AI and human actions share one audit trail. AI records also include the model and prompt versions used.
Human approval
When an AI-proposed action exceeds its autonomy level, a human reviews it before execution. Policies that require separation of duties can assign an independent approver.
- 1
Propose
The agent drafts an action — a message, a promise, a payment — tied to the case it serves.
- 2
Policy check
Current rule versions evaluate the action before it runs. Missing or stale information holds the action.
- 3
Approval if required
Only actions marked by task or policy rules wait for a person before execution.
- 4
Execute
The action uses the same controlled interface as human work. Financial figures come from ledger facts.
- 5
Audit record
Decision, model + prompt versions, approvals, and outcome land in a hash-chained record.
Learn & QA loopa quality program samples outcomes; findings tune prompts and policy without changing ledger facts
Task-level autonomy
Choose the right level for each task
Start each task type in shadow mode, then move it through summaries, assistance, human approval, or autonomous work as your evaluation results support.
Level 0
Shadow mode
Before a task type is activated, the agent runs in shadow: outputs are produced, recorded, and scored — never sent, never executed. Activation requires the evaluation gate to pass.
Level 1
Observe & summarize
The agent reads cases and produces summaries, timelines, and context briefs. It touches nothing.
Level 2
Assist
The agent drafts responses and recommends next actions. A human edits, owns, and sends everything.
Level 3
Human approval required
The agent prepares actions that policy marks for review, such as money movement or an exception, and a person approves each one before execution.
Level 4
Autonomous within policy
The agent executes on its own, inside hard guardrails — and still passes the compliance gate on every single action.
No task type skips shadow, and every level — including fully autonomous — still passes the compliance engine before each action and writes the same evidence record. Autonomy changes who clicks, never what is checked.
Ledger-backed facts
Keep borrower-facing financial figures ledger-backed
For configured borrower-facing templates, the platform fills financial figures from the ledger instead of asking the model to generate them. The model writes the words; the ledger supplies the numbers.
What the agent composes
Hi Jordan — as of {today}, your remaining balance is {ledger.balance} and your next payment of {ledger.next_due_amount} is due {ledger.next_due_date}.
Highlighted slots are resolved by the platform from the ledger at send time rather than generated by the model. A free-typed figure is held unless it resolves through an approved ledger-backed slot.
- Figures resolve at send time, so they reflect the ledger as of that moment — not as of model training or case open.
- The same mechanism covers balances, due dates, payoff amounts, and payment terms.
- If a required fact is missing or stale, the message is held until current information is available.
AI controls
Multiple controls around every agent
Ledger-backed facts and policy checks are supported by evaluation, version history, monitoring, a scoped kill switch, and human review.
Shadow mode before activation
A new agent type runs in shadow first: its outputs are recorded and scored, but never sent and never executed. It goes live only when the evaluation gate passes — not when someone feels confident.
Model and prompt version history
Each output records the model, prompt, and context version that produced it. New combinations must be registered and approved before activation.
Quality sampling & drift detection
The evaluation program uses per-case-type sample minimums and pass thresholds, then re-runs on model or prompt changes and on a regular cadence. It is designed to surface behavioral drift before it affects borrowers.
Fairness & consistency monitoring
Within the QA program, sampled outcomes are reviewed for consistency across borrower segments — and because every AI decision records its inputs, policy version, and model version, fair-lending review has the evidence it needs.
Instant kill switch
Suppress agents in one action, scoped to what is actually wrong: globally, per queue, per case type, or per channel. Human work continues uninterrupted.
PII scrubbed before model calls
Configured data policies remove or tokenize PII before a model call, so the model receives only the information needed for the task.
Human review by policy
Actions above an agent's autonomy level route to a human reviewer. Policies that require separation of duties can also require a different person from the action's human proposer.
Supported hosted or self-hosted models
Choose from supported frontier or self-hosted models while keeping the same policy checks, ledger-backed facts, version history, and audit trail.
Initial use cases
Start with scoped, repeatable servicing work
Begin with the tasks that consume your team today, using narrow permissions, clear handoffs, and the right autonomy level for each workflow.
Work collections calls end to end — hardship conversations included
AI voice agents place and answer collections calls: they negotiate payment plans and promises to pay within your guardrails, recognize hardship language and route it to the hardship workflow, and hand the call to a human the moment the borrower asks or policy requires. Like every task type, voice graduates — start with reminder and early-delinquency calls, and expand as your evidence supports it.
- Every dial clears the compliance gate first — configured contact-frequency thresholds, calling windows in the borrower's timezone, consent, and protected-borrower states.
- Every figure spoken comes straight from the ledger — the model speaks the words; the ledger supplies the numbers.
- Calls can be recorded and transcribed into the case timeline where permitted and configured.
Summarize cases
Start each case with a current brief: borrower context, account history, applicable protections, and amount owed.
Draft compliant responses
Replies drafted against the rules in force for that borrower, that product class, that jurisdiction — with every figure rendered from the ledger.
Recommend next actions
The agent recommends a next step and shows the policy check beside it for human review.
Offer payment plans within approved limits
Set the maximum term, minimum amount, and number of installments. Requests outside those limits route to a person.
Prepare payments for approval
Payment plans and transactions are staged with full context and routed for human approval before any money movement.
Your team stays in charge
Automation handles repeatable work; people remain responsible for decisions that require judgment. You set that boundary per task type and can change it as the evidence develops.
FAQ
The questions every risk committee asks
Yes. AI voice agents can handle collections calls, including hardship conversations. Before each dial, the platform checks configured contact-frequency limits, calling windows in the borrower's timezone, consent, and recorded protections. Financial figures come from the ledger, payment-plan terms stay within approved limits, and hardship language routes the case into the hardship workflow. The borrower can ask for a person at any point, and policy can require a handoff. Calls can be recorded and transcribed where permitted and configured. Voice starts with scoped use cases and expands as evaluation results support it.
Start with one scoped AI workflow
Begin in assist mode with your cases and policies. Review the proposals and controls, then expand autonomy only where your evidence supports it.