\t\tAccounting practice Case Study | Autoflow Solutions
← All case studies Back-Office Automation
Accounting practice / Italy

An accounting back office that reads the paperwork, files it, and reminds clients on time.

An Italian accounting practice spent its days keying tax forms, receipts, and bank statements by hand and chasing clients before payment deadlines. Autoflow built the back office that does it instead: documents are read and checked automatically, logged to the practice’s sheets, filed in Drive, and every client gets their tax payment reminder, with the right amount and form attached, on time.

Scope
  • Tax-form pipeline
  • Client reminders
  • VAT receipts
  • Statements to CSV
Stack
  • Make.com (orchestration)
  • PDF.co (document parsing)
  • OpenAI + Gemini (extraction)
  • Google Sheets + Drive

The challenge

The practice handles tax payments, VAT, and bookkeeping for companies and individuals. A large share of that work is the same steps on repeat: open a PDF, find the figures, type them into a sheet, file the document, and email the client before the deadline.

Done by hand, it kept two people busy, and every manual step was a chance to mistype an amount or send a reminder late. With tax payments, a late or wrong reminder is not a small slip: it lands on the client.

How the pieces connect

Read Tax forms, receipts, and statements are parsed and the figures extracted by AI.
Record Data lands in the practice’s Google Sheets, checked against what is already there.
Approve + file Staff mark each form approved or rejected; the PDF moves to the right Drive folder.
Remind Clients get the amount due and their forms by email at 10 days, under 10 days, and 1 day.

People approve and handle exceptions. Everything repetitive, the reading, the typing, the filing, and the reminding, runs on its own.

Autoflow’s role

Autoflow Solutions designed and built the full set of automations: the document-reading pipelines, the approval and filing flow, the scheduled client reminders, and the conversions into the formats the practice’s accounting tools need.

Constraints

  • Italian F24 tax forms come in more than one layout and belong to either a company or a person, so the extraction had to tell them apart rather than assume one format.
  • Amounts sent to clients had to be exactly right, including when one client has several forms due on the same date.
  • The practice already worked in Google Sheets and Drive, so the build had to fit their existing files and folders rather than move them to a new system.
  • A person still decides what is approved. The automation does the work around that decision, not the decision itself.

How we built it

  1. 01

    Teach the extraction the real document, not a generic one

    Each F24 is first sorted by layout, one page or two, and then by who it belongs to, a company or a person. Each combination has its own extraction prompt written around Italian tax-form structure, and the result is checked against the sheet first so the same form is never recorded twice.

  2. 02

    One reminder per client, with the total already worked out

    When a client has several forms due on the same date, the reminder adds them up, formats the total, and attaches every PDF to a single email. Reminders go out on a fixed schedule at 10 days, under 10 days, and 1 day before the deadline, and each send is recorded so nothing goes twice.

  3. 03

    Missing data stops the line and tells someone

    If a client has no email address on file, the reminder does not fail silently. The practice’s admin gets an alert naming the form and the deadline, so a person can step in before the date passes.

What we delivered

  • F24 tax-form pipeline: PDFs are read with PDF.co, the taxpayer, amounts, and due date extracted with OpenAI, checked for duplicates, logged to Google Sheets, and handed on to a Google Apps Script step.
  • An approve-or-reject flow: marking a form in the sheet moves the PDF to the matching Drive folder and updates its record.
  • Scheduled client reminders at 10 days, under 10 days, and 1 day before each payment date, with the total due and the forms attached, and an admin alert when a client email is missing.
  • VAT receipt processing for two client businesses: receipts dropped in Drive are read, sorted, and logged for VAT automatically.
  • Bank statements to CSV: statement PDFs are read with Gemini and converted into clean CSV files ready for import.
  • Document conversions: Meta ad invoices turned into XML, and a scanner that reads case documents with OCR (optical character recognition) and logs their type.

The results

2 fewer people needed for work the team used to do by hand
35% fewer errors, with no mistakes in reminder sends or tax amounts
0 late reminders to clients

Figures reported by the practice after the system replaced its manual process.

Evidence

The Make.com F24 extraction scenario, splitting by one-page or two-page forms and then by company or person.
Every F24 is read the way it is actually laid out. The image shows the Make.com extraction scenario: a duplicate check first, then a split by one-page or two-page form, then by company or person, each with its own AI extraction before the data is logged to the sheet.
The F24 tracking sheet with due dates, amounts, approval status, and sent-email ticks. Client details are hidden.
The practice works from one sheet the automation fills in. The image shows the F24 tracking sheet: due date, amount, approval status, and a tick once the client email has gone out. Client names, tax codes, and notes are hidden.
The Make.com reminder scenario that totals amounts per client, attaches the forms, and emails the client or alerts the admin.
One reminder per client, with the sum already done. The image shows the scheduled reminder scenario: it finds forms due, adds up each client’s total, attaches every PDF, emails one or two client addresses, records the send, and alerts the admin if an address is missing.
The Make.com VAT receipts scenario: Drive folder watch, PDF conversion, AI extraction, filing, and Google Sheets logging.
Receipts for VAT log themselves. The image shows the scenario for one client business: new receipts in Drive are converted, read by AI, filed into the right monthly folder, and added to that month’s sheet, with retries if a step fails.
The Make.com bank statement scenario converting a PDF statement into a CSV file in Drive.
Bank statements become import-ready files. The image shows the scenario that splits a statement PDF, has Gemini extract the transactions, cleans them with code, and saves a CSV back to Drive.

Want a system your business actually owns?

Tell us what breaks or slows you down. If a build makes sense, we will tell you. If it does not, we will tell you that too.

See other work