INVOICE PROCESSING

Agentic resolver for mismatched vendor invoices

When a PO match fails, an agent investigates the discrepancy across BigQuery PO and receipt data, drafts a clarification email to the vendor in Outlook.

CategoryInvoice Processing
Enginepaperclip
Difficultyadvanced
Triggerevent
Steps5
Setup~25 min

How it runs

The automated pipeline, trigger to output.

  • TriggerFlagged invoice mismatch event received
  • ActionInvestigate PO, invoices, and receipts in BigQueryGoogle BigQueryBigQuery
  • LogicDiagnose price, quantity, partial-ship, or missing-receipt cause
  • ActionDraft and send vendor clarification email via OutlookOutlook
  • OutputOpen Asana task tracking the exception to closureAsanaAsana

What it does

Takes a flagged invoice mismatch and does the legwork an AP clerk would: it digs into the PO history and goods-receipt records, forms a hypothesis for why the line does not match, drafts a specific question to the vendor, and tracks the open item until it resolves.

When to use it

Use it downstream of a matching flow when your exception queue is full of mismatches that each need investigation and a vendor conversation. It is agent-driven because the resolution path differs per case rather than following fixed branches.

How it works

  1. 1A flagged invoice mismatch event triggers the agent.
  2. 2The agent queries BigQuery for the PO, prior invoices, and goods-receipt records to understand the gap.
  3. 3It reasons about whether the issue is price, quantity, a partial shipment, or a missing receipt.
  4. 4It drafts a tailored clarification email to the vendor and sends it from the AP Outlook mailbox.
  5. 5It opens an Asana task with the diagnosis, the email sent, and the suspected root cause for AP to track to closure.

Set it up

What you configure once, before turning it on.

  1. 1
    Connect BigQueryDatasets, queries, schemas.
  2. 2
    Connect OutlookMail, calendar, contacts.
  3. 3
    Connect AsanaTasks, projects, milestones — everywhere.
  4. 4
    Set each agent's modelWe leave models unset so you pick the tier — fast + cheap, or top-quality.
  5. 5
    Tune it to your dataEdit the prompts, filters, and field mappings so it matches how your team works.
  6. 6
    Test, then turn it onRun once against a sample, confirm the output, then enable the trigger.

Run this workflow in your colony.

14-day trial. No DevOps. No Sales call. Provisioned in under a minute.