§ CASE 02 — INVOICE PROCESSING
Invoice processing pipeline
From PDFs in a shared inbox to fully-tagged Monday.com records with zero human touch.
Client identities protected under NDA. Reference calls available as part of serious engagement conversations.
§ THE PROBLEM
Every invoice was a human interrupt.
Every supplier invoice arrived as a PDF email attachment. The ops team had to open each one, read line items, identify invoice number and supplier, then manually create a Monday.com item with all fields populated. Each invoice took 8–12 minutes.
Errors were frequent because tab-switching between PDF and Monday.com is mentally exhausting work. The team was spending nearly 30 hours a month on what was effectively data entry — work that was never going to make anyone better at their job.
§ THE BUILD
A Make.com pipeline with Claude doing the reading.
We built a Make.com pipeline triggered by new emails in the supplier invoice inbox. The pipeline downloads the PDF, sends it to the Claude API with a structured extraction prompt, and parses the response into clean JSON. The structured data flows into a Monday.com "Create Item" module that creates a fully-tagged invoice record: supplier, invoice number, line items, totals, due date.
The whole pipeline runs in under 30 seconds per invoice. Edge cases (multi-page invoices, foreign currencies) handled with conditional routing and a fallback queue for human review.
§ OUTCOME
30 hrs/mo
Thirty hours a month returned to the ops team. Invoices now move from inbox to Monday.com with no human in the loop. The team that used to spend most of their week on invoice data entry now spends that time on supplier relationship work and AP reconciliation — the work the data entry was preventing.
§ THE SHIFT
The shared inbox stopped being a queue of dread.
The ops team's mornings stopped starting with "how many invoices did we get overnight?" Tab-switching between PDF and Monday.com — the most exhausting part of the work — disappeared entirely. The team noticed they were less tired by Friday. Errors that used to come from fatigue stopped showing up in the audit trail.
§ TECHNICAL DETAIL
Claude reads, Make routes, Monday records.
The Claude API call uses a structured-extraction prompt with a strict JSON schema, so the response is always parseable. Make.com handles the orchestration — Gmail watcher, PDF download, API call, JSON parse, Monday.com Create Item. Errors route to a dead-letter board so nothing silently disappears.
- OrchestrationMake.com
- Email triggerGmail watcher
- Document AIClaude API
- Record creationMonday.com Create Item
Want a system like this?
Every engagement starts with a 30-minute audit call. Tell us what's broken and we'll tell you whether we can help.