Request loans through AI chat

Workflow purpose

This guide demonstrates how to use Proto's platform to let a customer apply for a loan and get a pre-approval decision through AI chat, with manual review only for cases the credit engine can't clear automatically. Loan requests are transactional and repetitive by nature – amount, purpose, term – which makes them a good fit for automation, provided the actual credit decision stays with a proper credit engine rather than the AI agent itself.

Who can benefit from this guide:

Digital banks and microlenders offering AI-assisted loan applications
Lending teams wanting faster turnaround on straightforward applications
Operations teams reducing manual data entry on loan intake
Compliance teams needing consistent terms disclosure on every approval

Request a loan

This workflow captures loan details, checks eligibility with the credit engine, and either delivers terms, a decline notice, or a handoff for manual review. See how to request loans through AI chat.

Just starting with the platform?

For initial setup of your AI agent – including workspace setup, agent training, and channel deployment – please start here. This workflow guide assumes your AI agent is set up and ready for more advanced configurations.

1. Create a trigger

Purpose: Start the flow when a customer wants to apply for a loan.

Navigate to the Actions tab in the AI agent settings.
Click “+ Add Trigger” and select the Message Received trigger type.
Name the trigger “Loan Request”.
Add a description such as “Triggers if user requests to apply for or ask about a loan.”

2. Survey – loan details

Purpose: Capture what the credit engine needs to make a decision.

Add a Survey action titled Loan application.
Field loan_amount – Text – How much would you like to borrow?
Field loan_purpose – Quick replies: Business, Personal, Education, Home improvement.
Field loan_term – Text – Over how many months would you like to repay it?

3. Send API request – credit engine

Purpose: Get an actual decision rather than an estimate the AI agent invents.

Add a Send API request action to your credit decisioning engine with {loan_amount}, {loan_purpose}, and {loan_term}. Map the response to credit_decision.

4. Branch – branch on the decision

Purpose: Route each outcome to the right next step.

Add a Branch action.
Condition: credit_decision["status"] == "approved" → deliver terms.
Condition: credit_decision["status"] == "declined" → deliver a decline notice.
Condition: credit_decision["status"] == "review" → AI Agent Network handoff to the loan officer’s AI agent, passing {loan_amount}, {loan_purpose}, {loan_term}, and {credit_decision}.

5. Deliver the outcome

Purpose: Give the customer clear terms or a clear reason, either way.

Add a Send message action for the approved branch: “You’re approved for {loan_amount} over {loan_term} months at {credit_decision[“interest_rate”]}% interest. Full terms are attached – let us know if you’d like to proceed.“ Disbursement and acceptance are handled in a separate flow, once the customer confirms.
Add a Send message action for the declined branch: “We’re not able to approve this loan at this time. {credit_decision[“decline_reason”]}“

6. Test the flow

Purpose: Confirm every branch delivers the right message and nothing is left ambiguous.

Test an application that should auto-approve, and confirm the terms match what the credit engine actually returned.
Test an application that should be declined, and confirm the decline reason is included rather than a generic rejection.
Test an application that should route to manual review, and confirm the loan officer receives the full application details.

550K