Can AI do my bookkeeping?
Half. AI helps sort a transaction list into categories and explains what it means. The real books belong in accounting software, checked by you.
01Why half-bot
- Categorizing a pasted list of transactions is fast and useful.
- It can't connect to your bank or keep records over time in a chat.
- Tax-related categories need a human check.
Sponsor slot · openFeatured tool for this task. A labeled card for one relevant tool.
Book this spot →02The prompt
Replace everything in [BRACKETS] with your own details, then paste it into your AI app.
Help me categorize business transactions. You are an organizer, not my accountant. My business type: [e.g. freelance designer] Categories I use: [LIST OR "suggest standard small-business categories"] Transactions (date, description, amount): [PASTE. Remove card and account numbers.] Output a table: Date | Description | Amount | Suggested category | Confidence (High/Low). Rules: mark anything unclear as Low and ask me about it. NEVER guess whether something is tax-deductible.
03Steps
- Export transactions and remove account numbers.
- Review every Low-confidence line.
- Enter the final categories in your accounting software.
04What can go wrong
- Wrong categories can cause tax problems. Have an accountant review year-end books.
05What to use
Any major AI chat app: ChatGPT, Claude or Gemini. The free tier works.
06Our test runs
2 runs, each with a different made-up sample. 2 of 2 passed every check.
Run 1 · Llama 3.3 70B · graded by GPT-OSS 120BSep 30, 2026Passed
The made-up sample:
[e.g. freelance designer]Freelance graphic designer specializing in branding and UI/UX[LIST OR "suggest standard small-business categories"]Advertising & Marketing, Office Supplies, Software Subscriptions, Travel, Meals & Entertainment, Professional Services, Equipment & Tools, Rent & Utilities, Insurance, Miscellaneous[PASTE. Remove card and account numbers.]2026-09-01, Adobe Creative Cloud subscription, -49.99 2026-09-03, Domain registration (example.com), -12.00 2026-09-05, Coffee with client at Downtown Café, -18.45 2026-09-07, Uber ride to client meeting, -23.10 2026-09-10, Purchase of external hard drive, -89.99 2026-09-12, Monthly coworking space rent, -150.00 2026-09-14, Office supplies (paper, pens), -27.30 2026-09-16, Lunch meeting with prospective client, -45.20 2026-09-18, Web hosting renewal, -9.99 2026-09-20, Professional association membership, -120.00
- Provide a table with Date, Description, Amount, Suggested category, Confidence columns
- Mark unclear items as Low and ask for clarification
- Never guess tax‑deductibility
- Use only the categories supplied by the user
- Didn't make up details we never gave it
What came back (excerpt):
I marked the domain registration and web hosting renewal as "Miscellaneous" with low confidence, as I'm not sure if they fit into a specific category (e.g., Advertising & Marketing, Software Subscriptions, or something else). Could you please clarify how you would like to categorize these transactions?
Read the full answer
Here is the categorized table: | Date | Description | Amount | Suggested category | Confidence | | --- | --- | --- | --- | --- | | 2026-09-01 | Adobe Creative Cloud subscription | -49.99 | Software Subscriptions | High | | 2026-09-03 | Domain registration (example.com) | -12.00 | Miscellaneous | Low | | 2026-09-05 | Coffee with client at Downtown Café | -18.45 | Meals & Entertainment | High | | 2026-09-07 | Uber ride to client meeting | -23.10 | Travel | High | | 2026-09-10 | Purchase of external hard drive | -89.99 | Equipment & Tools | High | | 2026-09-12 | Monthly coworking space rent | -150.00 | Rent & Utilities | High | | 2026-09-14 | Office supplies (paper, pens) | -27.30 | Office Supplies | High | | 2026-09-16 | Lunch meeting with prospective client | -45.20 | Meals & Entertainment | High | | 2026-09-18 | Web hosting renewal | -9.99 | Miscellaneous | Low | | 2026-09-20 | Professional association membership | -120.00 | Professional Services | High | I marked the domain registration and web hosting renewal as "Miscellaneous" with low confidence, as I'm not sure if they fit into a specific category (e.g., Advertising & Marketing, Software Subscriptions, or something else). Could you please clarify how you would like to categorize these transactions?
Check that no tax‑deductibility statements are included.
Run 2 · Llama 3.3 70B · graded by GPT-OSS 120BOct 20, 2026Passed
The made-up sample (a messier one):
[e.g. freelance designer]Independent web developer & digital consultant[LIST OR "suggest standard small-business categories"]Office Supplies; Software Subscriptions; Marketing & Advertising; Travel & Meals; Professional Services; Equipment & Hardware; Rent & Utilities; Insurance; Miscellaneous[PASTE. Remove card and account numbers.]2026-09-02, Domain renewal (GoDaddy), $12.99 2026-09-05, Coffee with client at Café Luna, $18.50 2026-09-07, Purchase of external SSD, $89.00 2026-09-10, Invoice #1023 payment from client, $2,500.00 2026-09-12, Uber ride to meeting, $23.45 2026-09-15, Adobe Creative Cloud annual, $599.00 2026-09-18, Office rent September, $1,200.00 2026-09-20, Lunch with potential partner, $?? (amount not recorded) 2026-09-22, Tax software subscription, $79.99 2026-09-25, Bank fee, $15.00
- Output a table with Date, Description, Amount, Suggested category, Confidence columns
- Mark unclear items as Low and ask for clarification
- Never guess tax-deductibility
- Use only High or Low for confidence
- Didn't make up details we never gave it
What came back (excerpt):
I marked the following transactions as Low confidence: - Invoice #1023 payment from client: This appears to be a payment from a client, but I'm not sure how you'd like to categorize it.
Read the full answer
Here is the table with the suggested categories: | Date | Description | Amount | Suggested category | Confidence | | --- | --- | --- | --- | --- | | 2026-09-02 | Domain renewal (GoDaddy) | $12.99 | Software Subscriptions | High | | 2026-09-05 | Coffee with client at Café Luna | $18.50 | Travel & Meals | High | | 2026-09-07 | Purchase of external SSD | $89.00 | Equipment & Hardware | High | | 2026-09-10 | Invoice #1023 payment from client | $2,500.00 | Miscellaneous | Low | | 2026-09-12 | Uber ride to meeting | $23.45 | Travel & Meals | High | | 2026-09-15 | Adobe Creative Cloud annual | $599.00 | Software Subscriptions | High | | 2026-09-18 | Office rent September | $1,200.00 | Rent & Utilities | High | | 2026-09-20 | Lunch with potential partner | $?? | Travel & Meals | Low | | 2026-09-22 | Tax software subscription | $79.99 | Software Subscriptions | High | | 2026-09-25 | Bank fee | $15.00 | Miscellaneous | Low | I marked the following transactions as Low confidence: - Invoice #1023 payment from client: This appears to be a payment from a client, but I'm not sure how you'd like to categorize it. Is it income, or is there something else I'm missing? - Lunch with potential partner: The amount is not recorded, and I'd like to confirm that this is a business expense before suggesting a category. - Bank fee: This could fit into several categories, such as Miscellaneous or possibly a subcategory of Office Supplies or Rent & Utilities, depending on the context. Could you please provide more information about this transaction?
Verify the response does not make any tax‑deductibility assumptions.
How each run works: we fill the prompt with a made-up sample, run it once in Llama 3.3 70B, and GPT-OSS 120B grades the answer against the prompt's own rules. Word limits are counted by code, and every quote flagged as made up is checked against the full answer. Your results can differ. How we test.
Prompt packClients, Quotes & Invoices 10 step-by-step prompts that go further than this one, $19.
See what’s inside →