AI Bookkeeper

Pulls invoices and receipts from your inbox into a clean sheet, then emails you a weekly summary. Role: Bookkeeper / Ops Assistant.

Example shift, AI Bookkeeper · drafts wait for your reviewSynthetic example data

07:30
Invoice from Acme Hosting (PDF)Date, supplier, net, VAT, total, category extracted
Logged
07:31
Receipt photo, low confidence on the totalFlagged needs_review for you to check
Needs you
07:32
Expenses sheet: 14 rows addedMaths checked: subtotal + VAT = total
Logged
07:33
Invoice in USDKept in USD and totalled separately from EUR
Logged
Mon 08:00
Weekly summary emailed to youTotals by currency, top suppliers, duplicates left out of totals, rows to review
Draft ready

Example week at “growing company” volume: ≈ 5.4 h of first-pass work taken over (estimate)Open the setup guide

Implementation in your business

On the inbox where invoices and receipts arrive: it reads attachments, extracts the fields into an expenses sheet and sends you a weekly summary.

Where it plugs in

Invoice and receipt extraction (PDF and photo) from Gmail into an Expenses sheet with checks, plus a weekly finance digest.

What it needs

n8n (Cloud or self-hosted)
Google account: Gmail + Sheets
OpenAI API key
Your categories and home currency (Config)

What it outputs

One expenses row per document
needs_review flags on low confidence
Weekly spend summary email to you

Time saved (estimate)

≈ 23 h

per month at 120 documents a week, 4 min by hand vs 1 min review. See assumptions.

Setup: about 15 minutes on n8n Cloud

Three steps, no code. The setup guide has the full checklist.

Step 1 · 1 min

Pick the AI Bookkeeper

Start the trial on Starter (this employee) or Workforce (all five). The download link and guide arrive by email.

Plans

Step 2 · 3 min

Import it

n8n Cloud or self-hosted: Workflows → Import from file. No n8n? Paste the instructions into a Custom GPT or Claude Project.

2 workflow files
Sheet templates (CSV)

Step 3 · 10 min

Connect your tools

Google Sheets
OpenAI
Gmail
Config: your details and rules

Full setup checklist

How it works

Import-tested and test-executed with stubbed services on n8n 2.44 before release.

PDFs and photos. PDF invoices are read as text; receipt photos and screenshots go to the vision model.

Checks, not trust. Subtotal + tax must equal total, low-confidence rows and non-invoices are flagged needs_review, and possible duplicates (same vendor + invoice number) are listed in the weekly email.

Numbers come from code. The weekly totals are calculated in a code node; the AI only writes the narrative around them.

What you get

ai-bookkeeper-capture.n8n.json: Capture: Gmail emails with attachments → one item per PDF/image → PDF text extraction or vision → LLM fields (vendor, tax ID, invoice no., dates, subtotal, VAT, total, currency, category) → maths check → Expenses tab → label processed.
ai-bookkeeper-weekly-summary.n8n.json: Weekly Summary: every Monday 08:00 → Expenses tab → totals computed in code (by currency, category, vendor; unpaid; duplicates; needs review) → LLM writes the digest → emailed to you.
No-code version: Custom GPT / Claude Project instructions
Prompt library, sheet templates and a PDF setup guide

What it's worth to you

The value is time: the employee does the first pass and you keep the review. Two worked examples, every input an assumption:

Estimates from the assumptions in the table, not measured results. Coverage = documents the agent extracts cleanly (assumed 90%; the rest are flagged needs_review and counted as zero savings). The weekly summary's time saving is not counted. Not tax or accounting advice.
Assumption (estimate)	Freelancer / small team	Growing company
Invoices/receipts per week	30	120
Manual minutes per document	3	4
Review minutes (you still review)	0.5	1
Share the employee handles	90%	90%
Hourly cost of whoever does it	$25	$35
AI running cost per document	$0.02	$0.02
Hours taken over per week	1.1 h	5.4 h
Value of that time per month, after AI costs	≈ $120	≈ $810
After the $19 Starter plan	≈ $101	≈ $791

Try your own numbers

AI employee

Items per week

Minutes per item, done by hand Time yourself on 3 real examples.

Hourly cost of whoever does it ($)

More assumptions

Minutes to review each output You still review everything.

Share the employee handles (%) e.g. support emails answerable from your knowledge base.

AI running cost per item ($)

Plan cost per month ($) Starter is $19, Workforce $49 for all five.

Hours taken over per week-

Hours per month-

Value of that time per month-

Net per month after AI costs and plan-

Hours saved needed to cover the plan-

Formula: hours/week = volume × share handled × (manual − review minutes) ÷ 60 · hours/month = hours/week × 52 ÷ 12 · net = hours/month × hourly cost − AI running costs − plan cost. Time saved only; no claims about revenue, replies or hiring outcomes.

Questions

Does it connect to Holded, QuickBooks or Xero?

v1 writes to Google Sheets, which all of them can import. Native nodes exist for QuickBooks and Xero in n8n; Email support if you want pointers.

Does it handle multiple currencies?

Yes. Totals are grouped by currency; there's no FX conversion.

Do I need n8n?

For the automated version, yes: n8n Cloud or the free self-hosted edition. Every employee also ships a no-code version for a ChatGPT Custom GPT or a Claude Project.

Does it act on its own?

It works in drafts and logs. Anything that reaches a customer, candidate or prospect is a draft you review and send.

What does it cost?

Starter is $19/month for this employee, Workforce $49/month for all five. Both have a 7-day free trial. You also pay your own AI usage (usually cents per item) and your n8n plan if you use n8n Cloud.

Guardrails

This is data entry, not tax or accounting advice. Your accountant/gestor files taxes.
Spot-check totals before exporting. Never pay an invoice based only on extracted data (invoice fraud often arrives by email).
Financial documents go to your LLM provider. Check this fits your data policy.

Put the AI Bookkeeper to work.

Start 7-day free trial$19/mo for this employee, or $49/mo for all five.