Department automation program

Where it started

A mid-to-large fintech was growing faster than its processes. Registries, quotes, invoices and collections were spread across the CRM, email and each department's spreadsheets, and moving them from one place to another was manual work.

The challenge

Automate without losing control: every department had to know what runs, when, with what result and who to alert if something fails. And make it work for five departments with different needs, across thousands of invoices.

How I approached it

I started with a pilot in a single department: an n8n flow with Python that turned an Excel file into PDF documents. With that pattern proven, I built the portal and added departments one by one.

I coordinated several collaborators by task, with one pattern for everyone: every automation with its ID, trigger, log and owner, reviewed before it was switched on.

End to end

The flows run on n8n and Make, with Python scripts for the heavy lifting, and connect through REST APIs to the CRM, email and Microsoft 365. Data travels as JSON from one step to the next.

There are two kinds of automation. Automatic ones run on their own schedule: they're monitored, paused and ticketed. Manual ones are launched by someone and move step by step, with what each step needs.

What I built

A department portal with four roles—admin, leadership, department head and visitor—and Microsoft 365 sign-in.

• Registry matching into PDF formats.
• Quote labeling.
• Payment reminders across thousands of invoices.
• Weekly accounts-receivable summary.
• Access by ticket: other departments' folders ask for credentials and every access goes into the log.
• Failure, improvement or observation tickets, with the reply sent to whoever raised them.
• Audit: runs, attention needed, failures, and weekly and monthly reports; alerts arrive by email and in the CRM.

What changed

Each department sees its processes running 24/7, with their results and history. What fails raises an alert and a ticket, and what gets fixed is measured again.

What I learned

That an automation without visibility is a black box: the log and the alerts matter as much as the flow. And that leading other developers starts with a clear, proven pattern everyone can follow.