AI agents

Agent Builder

Every lender has processes that don't fit a product roadmap. With the Agent Builder you describe the process in plain English. AI drafts it as a decision tree, explains each step back to you, and your team reviews and switches it on.

Example: a loan falls into arrears

A collections agent built with Agent Builder: it proposes the next step, a collector approves it.

The agent recommends; a collector approves. Hardship and payment arrangements always come before any notice.

Ask it in plain English

On any loan, a collector can ask the agent what is happening and what to do next. It answers from the loan’s own records, shows its reasoning, and acts only when you approve.

AI

Collections agent

On loan 1000003104 · example with sample data

What's the status of this loan?

Car loan 1000003104 for Sarah Mitchell.

  • Balance $38,412.60
  • One repayment of $812.40 missed: 34 days past due
  • No hardship or payment arrangement in place

When was the last contact made?

The 5-day late letter was emailed on 5 September, 29 days ago. Nothing since.

It's in arrears. What should we do?

34 days past due with no contact this cycle, and no hardship or arrangement to respect. I recommend sending the pre-default reminder.

Approved

Done. The pre-default reminder was emailed to Sarah and logged on the loan.

What actions have been taken on this loan?

Two so far:

  • 5 Sep: 5-day late letter, by email
  • Today: pre-default reminder, by email, approved by you

What the agent is doing

  1. ✓

    Read the loan

    34 days past due

  2. ✓

    Read the contact history

    2 actions

  3. ✓

    Check protections

    None in place

  4. ✓

    Measure severity

    Moderate, 30 to 89 days

  5. ✓

    Choose the next step

    Pre-default reminder

  6. ✓

    Wait for your approval

    Approved by you

  7. ✓

    Send and log it

    Logged on the loan

Hardship and payment arrangements are always checked before anything is recommended, and nothing is sent without your approval.

An example conversation on a loan in arrears. Names and figures are sample data.

What it does

What it never decides

An agent is a set of rules you can read. AI helps you write it, but nothing goes live until your team approves it, and it can only do what you've allowed.

How we govern AI

The other agents

See it on your own process

Book a call