Credit File Completeness Agent
Asks for everything missing once, not one document at a time
- 9hard rules
- Things it must never do, enforced in every run
- 5handover rules
- Conditions where it stops and gives the work to a person
- 11output fields
- A structured result, not a paragraph you have to re-read
- ≤7steps per run
- Bounded work: at most 5 tool calls, enforced by the runtime
Read-only by design. This agent has no path to sending, spending, publishing or committing anything. Where that limit is the product, removing it would remove the reason to trust the output.
Overview
A lending decision is made once, but the file supporting it is assembled over weeks by people who cannot see the whole. So the same failure repeats at every lender: the file reaches credit, credit asks for one more document, the request goes back to the customer, and four days disappear. Then it happens again, because the second missing item was never noticed while asking about the first.
That is the largest controllable cost in lending operations, and it is not a credit problem. Every hour of it is spent on files that were always going to be approved.
The quieter cost is worse. A document that is present but stale — a valuation older than policy allows, expired identification, financials predating a required reporting period — passes a checklist because the box is ticked. That is the version that fails a file review, after the money has gone out.
This agent checks a file against your own policy for what is missing, expired or stale, and produces one complete request instead of four sequential ones. It is deliberately built to have no opinion on the borrower: no credit assessment, no score, no view on whether anyone should be lent to. Those require a regulated process with accountability it cannot carry.
Benefits
Days removed from every file
By asking once instead of four times.
Fewer file review findings
Because stale is caught before drawdown, not after.
Customers asked once
Which is most of what they remember about the process.
What it does
One request instead of four
Every missing item found at once. The sequential ask is the largest controllable delay in lending operations.
Catches present-but-stale documents
A valuation older than policy allows ticks the box and fails the file review. That failure lands after the money has gone.
Checks against your policy, not a generic list
Requirements differ by product, amount, security and borrower type. It reads yours.
Checks internal consistency
The same figure appearing differently in two documents — a query to resolve before submission.
Says what is conditional
Documents required only if something else is true, which is where generic checklists always fail.
Makes no credit assessment
No score, no view on the borrower, no recommendation to lend or decline. It checks the file, not the person.
What this changes
What this changes about your week
- It takes over
- Checking a lending file against policy for everything missing or stale, at once.
- Done today by
- Credit, one document at a time, four days apart.
- How often
- Every file, at submission and again at drawdown.
- Typical manual effort
- Requirements differ by product, amount, security and borrower, and half are conditional.
Why Banking & Financial Services in particular. The sequential ask is the largest controllable delay in lending, and it is spent on files that were always going to be approved. Worse, a present-but-stale document ticks the box and fails the file review after the money has gone.
Runs unattended
Started by you or by an event, and it finishes on its own. Nothing waits for someone to be at a desk.
The same standard every time
The two-hundredth item is held to the bar the first one was. Consistency is the part people cannot sustain.
It cannot act on its own
Credit File Completeness Agent has no path to sending, spending or committing. That limit is why its output is safe to act on.
How it runs
This agent runs server-side through the PROMIVO runtime. Each run is logged step by step and every tool call is permission-checked before it executes.
- 1InputValidated against the agent input schema
- 2Modelclaude-opus-5 with automatic fallback
- 3Toolssearch_knowledge, extract_structured_data
- 4PermissionsRuns READ actions automatically; anything higher waits for approval
- 5OutputStructured result matching the output schema
- 6LoggingTokens, duration and every step recorded
Demo
Demo dataIllustrative sample output, abridged.
Input
{
"amount": 1400000,
"product": "Commercial mortgage",
"documents": [],
"borrowerType": "Limited company",
"securityType": "Commercial property",
"fileReference": "FIL-20418",
"expectedDrawdown": "2026-10-15"
}Output
{
"missing": [
"Environmental screening report — required for commercial property security and not present in any form.",
"Identification for the second director."
],
"escalate": true,
"disclaimer": "A completeness and currency check against your own documentation policy. No credit assessment, score, rating or recommendation has been made; nothing has been said about the borrower, their affordability or the merits of the application; and no file has been submitted, approved or progressed.",
"policyFound": true,
"couldNotCheck": [
"The policy is silent on whether a screening report is required where a recent one exists for an adjoining site."
],
"presentButStale": [
{
"note": "Present, so it passes a checklist. At the expected drawdown date it will be 9 months old, outside policy. This is the finding that surfaces in a file review after the money has gone out.",
"dated": "2026-01-08",
"document": "Valuation",
"policyWindow": "6 months at drawdown"
}
],
"escalationReason": "A required document is absent at submission, the valuation will be outside the policy window at drawdown, and two documents state different turnover for the same period.",
"consistencyQueries": [
"Turnover is stated as 4.2m in the management accounts and 3.9m in the application form for the same period. A query to resolve, not a conclusion about anything."
],
"consolidatedRequest": [
"An environmental screening report for the security property.",
"Photographic identification for the second director.",
"A valuation dated within 6 months of the expected completion date.",
"A refreshed company search after 1 October.",
"Confirmation of which turnover figure is correct for the period."
],
"requirementsApplied": [
"Commercial mortgage over 1m: two years of audited financials rather than one.",
"Limited company borrower: directors' identification and company structure evidence.",
"Commercial property security: valuation within 6 months at drawdown, plus environmental screening."
],
"expiringBeforeDrawdown": [
"The company search expires 2026-10-01, before the expected drawdown. Refresh it with the other items rather than twice."
]
}Requirements
Integrations
No integrations required.
Before you start
- Your documentation policy by product, in a knowledge base
- The file's documents with their dates
- Facility details: product, amount, security, borrower type
- Requires the All-Access plan or higher
Use cases
Pre-submission check
Before a file goes to credit, not after it comes back.
Pre-drawdown check
Whether anything has gone stale since approval.
File review preparation
What a reviewer would raise, found first.
A new product or policy
Whether files are actually meeting the changed requirements.
Pricing
All-Access
Billed monthly through your PROMIVO subscription. Cancel at any time.
Runs consume your plan allowance for agent executions and tokens. See plan limits
FAQ
Does it make lending decisions?
No, in the strongest terms. It produces no credit assessment, no score, no view on affordability and no recommendation. It checks whether the file is complete and current against your policy.
Does it assess the borrower at all?
No. It never characterises a borrower, their circumstances or their creditworthiness. Where a document raises a question it reports it as a document query to be resolved by a person.
Can it approve or decline?
It cannot, and it has no route to any lending system. Approving credit is a regulated decision with accountability attached to a named person.
Reviews
Reviews open once customers have run this agent.
Need this agent to work differently?
Tell us what to change and our team will scope a customised version for your business.
Customize this agent