Workflow examples

AI document processing that prepares the file and leaves the decision to you

A loan, mortgage or credit application arrives as a pile: IDs, payslips, bank statements and contracts, in every format a phone can produce. Before anyone can decide, someone has to sort it, read it and check that nothing is missing. Here agents do that preparation and write a one-page summary — and your credit officer decides and submits, as before.

The workflow

When an applicant's document set is complete and approved, the agents sort and read every file, check the set against your checklist and write a one-page summary — then the summary and the complete file wait for your credit officer's approval.

  1. Trigger

    When someone messages you

    The applicant sends documents on WhatsApp — the ones you marked Sensitive, like IDs and bank statements, through the Secure upload link — and the workflow starts once the set is complete and you've approved it.

  2. Agent

    Classify

    Classify sorts each file by type: ID, payslip, bank statement, contract or other.

  3. Agent

    Agent

    The agent reads every document and checks the set against your checklist table: what's there, what's missing, what doesn't agree.

  4. Agent

    Artifacts

    An agent that works with files writes a one-page summary as a PDF and saves it in Artifacts.

  5. Approval

    Approval

    An Approval step puts the summary in front of your credit officer, by email or on Slack.

What it uses

WhatsApp number in Settings → Channels, Sensitive fields, Secure upload link, A form people fill in, Classify, Agent, Tables, Skills, Artifacts, Approval

Skill credit-checklist.mdloads every run

Which documents each product needs, what counts as consistent, and what the summary must cover

What does the agent do with the documents?

It prepares. Classify sorts each file by type. The agent then reads the payslips, the bank statements and the rest as an assistant would before a meeting: do the names, dates and employers agree across the ID, the payslips and the statements, and do they match the application? It checks the set against your checklist table, so a missing statement or a payslip from the wrong month is flagged before anyone opens the file — a first read, not a guarantee, and your credit officer still checks the set. An agent that works with files then writes a one-page summary as a PDF, saved in Artifacts.

Who makes the credit decision?

A person, always. The workflow doesn't make credit decisions. An Approval step puts the summary and the complete file in front of your credit officer, by email or on Slack, and the run waits for their decision. Your credit officer decides and submits the application to the bank or the portal themselves: neonloops doesn't log in to portals or submit applications.

How do the documents get in?

Most applicants send them on WhatsApp. Mark the fields for IDs, payslips and bank statements Sensitive and the agent never asks for them in the chat — Meta's rules don't let a business ask for ID or account numbers there. The applicant gets a Secure upload link instead: a private page where they agree and upload. The agent reads every other file as it arrives and asks again for anything blurry or missing, and a reminder template nudges an applicant who goes quiet. Nothing starts until the set is complete and you've approved it.

Documents can also come in through A form people fill in, a hosted page. Its file fields aren't the sensitive kind, though, so for IDs and statements the WhatsApp upload page is the better door.

What won't it do?

It doesn't make credit decisions, and it doesn't submit applications to a bank or a portal. It collects KYC papers but doesn't verify that an ID is genuine. It doesn't understand voice notes. neonloops is hosted in Helsinki, and the AI models that read the documents run with providers in the US, so check that this fits your own data protection obligations before you start.

What does it cost to run?

neonloops bills for the time your work is actually working: the minutes the agents spend sorting, reading and writing the summary. The days an applicant takes to send the last statement, and the hours the file waits for your credit officer, cost nothing. A workspace adds a card and is priced per billable minute from the start, against a ceiling you set yourself; there's no trial. Meta bills its own fees for WhatsApp templates such as reminders, to your business's own Meta account.

How pricing works
Start a workspaceBook a 30-min call