A landlord can use AI to track rental income and expenses each month by exporting the bank statement as a CSV, having an assistant suggest a property and category for every line, matching receipts, checking rent received against rent due, and saving a short pack for the accountant. Allow about 90 minutes once the routine is set up.
The assistant does the slow, repetitive reading: decoding bank descriptions, spotting which property a plumber's invoice belongs to, turning a photo of a receipt into a row. You still own the decisions that change the numbers, which are the category of unusual items, the repair-or-improvement call and anything involving the mortgage or deposits. Every step below keeps to that split.
It assumes one to ten properties and a person, not a bookkeeping department. If you already employ a bookkeeper, a bookkeeper's receipt capture and categorisation setup covers the same ground from their side. This is general information, not tax or accounting advice; your accountant decides how items are treated.
Set up the structure once so AI has something to fill
An assistant is only as useful as the structure you give it. Before month one, spend an hour on three things.
A short list of categories
Fewer is better. Fifteen categories that you use consistently beat forty you argue about. A workable list for residential lets:
- Rent received
- Other income (late fees, parking, laundry)
- Mortgage interest
- Repairs and maintenance
- Letting agent and management fees
- Insurance
- Utilities and local service charges you pay
- Ground rent and building service charges
- Safety certificates and inspections
- Advertising and tenant-finding costs
- Legal and professional fees
- Travel for property business
- Capital items (flagged for the accountant)
- Deposits held or returned (not income or expense)
- Personal or unrelated (to be excluded)
Ask your accountant to approve the list once. It then doubles as the handover format, which saves a tedious reformatting job at year end.

The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowA property key
One row per property with a short code, the address, the tenant's first name or unit number, the monthly rent due and the due day. The assistant uses the codes to tag every line, for example "12-OAK" or "FLAT-3". Without a key it will invent its own labels, and they will not match next month.
A separate account, if you can
Rental income and expenses going through a dedicated account make every later step easier. If you cannot separate them, you will need a "personal" category and the assistant will make more mistakes. A separate account is the cheapest accuracy improvement in this whole routine.
The monthly routine, step by step
| Step | What you do | Typical time |
|---|---|---|
| 1. Export | Download the month's statement as CSV; strip account numbers | 5 minutes |
| 2. Categorise | Assistant tags property and category; you resolve flagged lines | 20 minutes |
| 3. Capture receipts | Photograph or forward invoices; assistant extracts and matches | 20 minutes |
| 4. Reconcile rent | Compare rent received with rent due by property | 10 minutes |
| 5. Month-end check | Totals, unmatched items, unusual amounts, mortgage split | 15 minutes |
| 6. Save the pack | Ledger, receipts folder and a one-page summary | 10 minutes |
Step 1 and 2: statement upload and categorisation
Export the statement as a CSV rather than a PDF; columns survive, which keeps the output consistent. Open it and delete anything you would not want to see in a screenshot: account numbers, bank codes and personal transactions if the account is mixed. Then upload the file along with your category list and property key, in an assistant on a plan whose data settings you have checked. The question of connecting live bank access is separate, and is it safe to connect AI tools to your business bank account sets out the risks; for this routine, a manual CSV upload is enough.
I am a landlord with the properties in the attached key. Categorise every line of the attached bank CSV.
Return a table with columns: Date | Description | Amount | Property code | Category | Confidence (High/Medium/Low) | Question.
Rules:
- Use ONLY the categories and property codes provided. If unsure, use "Needs review" and put a short question in the last column.
- Rent received: match on amount and tenant name where possible.
- Do not guess between repair and capital improvement; mark "Needs review".
- Mortgage payments: category "Mortgage payment (split needed)".
- Do not combine or drop lines. The number of rows out must equal the number of rows in.
At the end give: total in, total out, number of Needs review lines.
The last rule, equal rows in and out, is a quiet guard against the assistant dropping lines when a file is long. After the table, compare row counts and the total of the statement against the bank's closing balance movement.
A sample extract of what comes back (illustrative):
03/10 | TFR IN RENT 12 OAK | +950.00 | 12-OAK | Rent received | High |
05/10 | ABC PLUMBING 4471 | -186.00 | 12-OAK | Repairs and maintenance| Medium | Invoice says boiler valve; confirm not part of replacement
08/10 | DD LANDLORD POLICY | -34.50 | ALL | Insurance | High | Same as last month
12/10 | AMZN MKTP 8X2Q | -61.99 | ? | Needs review | Low | Personal or rental item?
15/10 | MTG LENDER 99213 | -742.18 | ALL | Mortgage payment (split needed) | High |
What you would fix: the plumber line is the interesting one. The assistant tagged it as a repair at medium confidence and flagged a real doubt. You open the invoice, see that it is a valve replacement, and confirm. The Amazon line is where it is honestly unsure, and "Needs review" is the right output. The failure to watch for is the opposite: a low-quality tool that labels the Amazon line "Repairs and maintenance" with false confidence because it saw the word "marketplace".
Keep a "rules" list that grows each month: "DD LANDLORD POLICY = insurance, split across all properties". Paste it into the next month's prompt and the Needs review count drops. Note that count each month: it should fall as the rules list grows, and how far it falls depends on how regular your transactions are.
Step 3: receipt capture that matches the statement
Receipts and invoices are where records fall apart. Settle on one place they go: a shared folder, a dedicated email address that forwards, or the receipt feature in your ledger. Then each month give the assistant the photos or PDFs and ask for rows, not prose.
Extract from each attached receipt or invoice: Supplier | Invoice date | Invoice number | Net | Tax shown | Gross | Description | Property (from address or job notes, else "Unknown").
If any field is unreadable, write "UNREADABLE", do not estimate.
Then match each to a line in the bank table above by amount within 0.00 and date within 14 days. List unmatched receipts and unmatched payments separately.
Two lists result, and they are the useful bit. Unmatched payments are money that left the account without paperwork. Unmatched receipts are costs that have not yet been paid or hit a different account. Each should be empty or explained by month end. A receipt for "Unknown" property is a good example of what the assistant should not guess; look at the work order and ask the contractor if needed.
For a quick rule of thumb, photograph receipts the day you get them. A glove-box full of crumpled slips at month end is a slow way to work, and no assistant fixes a lost receipt.
Step 4: reconcile rent received against rent due
Categorising a rent payment is not the same as knowing that the right amount arrived on time. Use the property key's rent due and ask the assistant, or a simple spreadsheet formula, to compare.
| Property | Rent due | Received in month | Difference | Note |
|---|---|---|---|---|
| 12-OAK | $950 | $950 | $0 | Paid 3 Oct |
| FLAT-3 | $820 | $410 | -$410 | Part-payment, tenant message 9 Oct |
| 7-MILL | $1,100 | $1,100 | $0 | Paid early |
| Total | $2,870 | $2,460 | -$410 | One arrears item |
The arrears figure should lead to an action: a reminder, a conversation, or a note that a payment plan is agreed. Writing the reminder is a good small use of the assistant, provided the amount, the date and the tenant's name are copied from your table, not from the model.
Worked example: three properties, one October
An illustration with invented numbers. A landlord with three rentals (a terraced house, a flat and a small shop unit) exports October's statement: 31 lines. Rent due is $2,870; the statement shows $2,460 of rent received, plus a $200 refund from a supplier, against a $742.18 mortgage payment, a $186 plumbing invoice, $34.50 of insurance, a $61.99 unidentified card payment and several smaller items.
The assistant's table shows 24 lines tagged High, 5 Medium and 2 Needs review. The landlord spends 12 minutes on the 7 uncertain lines: confirms the plumbing is a repair, identifies the card payment as a smoke-alarm battery pack for the flat (also a repair) and marks one $35 payment as personal. The receipt step extracts five invoices; four match a payment and one electrician's invoice for $310 has no matching bank line because it was paid in cash. That is a real finding: the cost is genuine but the paper trail is thin, so the landlord asks the electrician for a receipt and resolves to pay by transfer next time.
The rent reconciliation shows $410 outstanding on the flat. The month-end check finds that the mortgage line needs splitting using the lender's statement; in this example, $448 is interest and $294.18 is capital repayment. The interest goes on the mortgage-interest line and the repayment reduces the loan balance, because the two are treated differently; how the interest counts for tax is your accountant's call. The summary page lists rent received, expenses by category, the arrears item and three open questions for the accountant. The shape of the month (a thin result because of an arrears item and a one-off repair) is ordinary, and it is exactly why a monthly review pays off: nobody wants to discover a missing $410 at year end.
Total time: about 85 minutes the first month and under an hour by the third, since the rules list and property key now handle most lines.
Step 5: the month-end check
This is the step that decides whether the books can be trusted. Run it every month, using a checklist rather than memory.
- Row count and totals. Rows in the table equal rows in the statement; money in minus money out equals the account's balance change.
- No "Needs review" left. Each has been resolved, or is parked with a dated question for the accountant.
- Unmatched receipts and payments. Both lists are empty or each item has a one-line explanation.
- Odd amounts. Sort by amount, look at the top five, and ask whether each makes sense. A $5,000 payment tagged "Repairs" deserves a second look.
- Same supplier, different category. Ask the assistant for suppliers tagged to more than one category this month. Sometimes it is correct; often it is a slip.
- Mortgage split recorded from the lender statement.
- Deposits. Any deposit received or returned is kept out of income and expense, and matches your deposit protection records.
- Rent roll. Every property's rent is either received or noted as arrears.
Ask the assistant to run checks 4 and 5 as a "second pair of eyes" in a fresh chat with only the table. It is cheap and occasionally catches a transposed digit. It cannot replace checks 1 and 3, which depend on facts outside the table.
Questions to ask your own ledger afterwards
Keep each question narrow. "How is my portfolio doing?" produces a paragraph of reassurance; "which property cost the most per dollar of rent?" produces a number you can check and act on.
Once the month is clean, the table is a small database you can question in plain English. This is where an assistant earns its place beyond data entry, and where a landlord starts to see the business and not only the paperwork. Paste the table and ask one question at a time.
Using only the attached table, answer:
1. Total repairs and maintenance by property this month.
2. Which property has the highest cost as a share of its rent?
3. Any supplier I paid more than once this month?
4. Compare this month with the previous month's table for insurance and utilities.
Show the arithmetic for each answer.
A typical reply (illustrative): "Repairs: 12-OAK $186.00, FLAT-3 $61.99, 7-MILL $0.00. FLAT-3 costs are $61.99 against $820 rent, 7.6%; 12-OAK $186 against $950, 19.6%. No supplier appears twice. Insurance is unchanged at $34.50." What you would check: the percentages. Divide $186 by $950 yourself; it is 19.6%, so the answer stands. A habit of verifying one figure per answer is quick and catches the occasional slip.
These questions turn into decisions. If one house regularly takes a fifth of its rent in repairs, that points at the boiler, the roof or the tenant mix, and a conversation with the managing agent or a quote for planned work is cheaper than another year of call-outs. For the repair side, see how AI helps with tenant messages and repairs.
Slips that happen in this routine, and how they show up
- A duplicated import. Two statements overlap by a few days and the same rent payment is counted twice. It shows up when rent received exceeds rent due. Fix: filter by date to the exact month before uploading.
- Rent and deposit mixed. A $1,500 deposit lands in the same account and is tagged as rent. It shows up as a rent figure far above due. Fix: a rule that any amount equal to one month's rent plus a deposit-type reference goes to "Deposits held".
- A split payment. A tenant pays $410 twice in one month. The assistant tags both as rent received for the flat, which is right, but your comparison shows a difference only if you add both. Fix: reconcile by summing per property, not per line.
- A supplier serving several properties. One insurance direct debit covers all three; tagging it to one house distorts the cost per property. Fix: tag "ALL" and split evenly or by rent at month end.
- A changed description. A bank starts printing "TFR IN" differently and the rule list stops matching. It shows up as a jump in Needs review lines. Fix: update the rule, do not accept the lazy label.
What to flag for the accountant, not decide yourself
A good routine makes the accountant's work easier and cheaper. Keep a running "questions" list with the transaction, the date and what you think; do not let the assistant settle these.
- Repair or improvement. A new boiler is not the same as a fixed valve. Describe the job, the cost and what it replaced.
- Furniture, appliances and equipment. Treatment depends on your rules and whether the property is let furnished.
- Costs before a property was let. Work done while the property was empty may be treated differently.
- Mixed-use payments. Travel, phone, home office; only a fair business share counts.
- Anything with a large amount or unfamiliar nature. Ask first, record the answer, and add it to the rules list for next time.
If you want to understand where AI bookkeeping tends to be wrong in general, can AI do my bookkeeping and what still needs an accountant covers it, and AI bookkeeping software versus a human bookkeeper compares accuracy and cost.
Handover pack for the accountant
At year end, or whenever your accountant asks, a clean pack saves them time and you money. Each month, save these into a dated folder, such as "2026-10":
- The categorised ledger as a spreadsheet (one tab per month or one running tab with a month column).
- The original bank CSV, untouched.
- The receipts and invoices folder, named by date and supplier.
- The one-page summary: income and expenses by property, arrears, open questions.
- Lender statements for mortgage interest and any loan changes.
- Rules list and property key, with the version date.
The summary is a good job for the assistant, provided the numbers come from the ledger. Give it the table and ask for a half-page narrative with totals by property and a list of open questions, then check each total against the spreadsheet. If you want a ledger underneath, a free tier can be enough: Stessa's free Essentials plan lists automatic bank feeds, document storage and basic financial reports, with receipt scanning capped at five a month, and the paid Manage and Pro plans add more reports; prices and limits change, so check the vendor's page before choosing. Tools such as Xero or QuickBooks are the heavier route if you also run the lets through a company.
One last habit: once a quarter, ask the accountant to look at one month's pack and tell you what they would change. Two sentences of feedback improves the category list more than hours of tweaking the prompt, and it keeps the routine something both of you trust.
For help turning this into a routine that fits your own portfolio, that is what my AI implementation consultation covers.
Practical questions landlords ask about AI bookkeeping
Can I just give ChatGPT or Claude my bank statements?
You can upload a CSV export, but remove account numbers and any personal transactions first, and use a plan whose data settings you have checked. Treat the output as a draft: the assistant suggests categories, you confirm them. It does not replace a ledger or bookkeeping software.
Do I need rental bookkeeping software as well as AI?
A spreadsheet is enough for one to three properties. Beyond that, software that pulls bank feeds, stores receipts and produces property-level reports saves time. AI then handles the awkward parts: odd descriptions, receipts and your questions. Pick the ledger first and add AI on top.
How do I split mortgage payments between interest and repayment?
Take the figures from the lender's statement, not the bank line, because interest and capital repayment are treated differently: the repayment reduces the loan and is not a running cost, while how the interest counts for tax depends on your rules. Record the two separately, ask your accountant how to treat them, and give them the lender's annual statement.
What should I never let the assistant decide?
Whether a cost is a repair or an improvement, how to treat deposits, and anything that affects tax. Let it flag candidates and explain the question, then ask your accountant. A wrong guess here changes your tax figures, and the assistant cannot see your full circumstances.
Further reads
- How to Choose an Outsourced Bookkeeping Service That Uses AI — If you would rather hand the routine to a bookkeeper.
- Is AI Worth It for a One-Person Bookkeeping Business? — The same question from a one-person bookkeeper's side.
- AI for Rental Properties: Screening Enquiries, Viewings and Tenant Replies — Filling vacancies, which is where rent income starts.
- How Much Does AI Bookkeeping Software Cost per Month? — What bookkeeping software costs if you outgrow a spreadsheet.
- AI Assistant for Landlords: WhatsApp Maintenance Requests and FAQs — Maintenance requests that generate the invoices you will book.
- AI Tools for Landlords: Messages, Lease Checks and Application Fraud — Tenant messages, lease checks and application fraud.
- AI Software for Landlords: Rent, Maintenance and Reminders Compared — TenantCloud, TurboTenant, RentRedi, Innago, Baselane and Buildium compared on price per unit, free tiers and what the AI really does.
- Best AI Property Management Software in the UK: A Buyer's Checklist — Latch, August, Landlord Studio, Street and Arthur compared for a small lettings portfolio, with tests for the AI claims and the exit.
- AI Real Estate Investment Analysis: Vet a Rental Deal Step by Step — A rental-deal checklist for small investors: rent comps, assumptions table, cash flow, a seller cap rate check and the AI maths errors to catch.
- AI Rent Roll Analysis for Multifamily: Reconcile to the Bank — Clean a rent roll, summarise vacancy and delinquency, and tie billed rent to bank deposits with AI, using four prompts and a worked 48-unit example.
- Making Tax Digital for Landlords: Keep Rental Income Records With AI — A landlord's quarter-by-quarter record routine with a filled ledger, AI checks and software submission gate.
- Property Investment AI Tools: A Stack for a Five-Property Portfolio — A five-property investor can start with free deal and bookkeeping tools, then pay only for the workflow that repeatedly saves review time.
- Using AI for Property Investment: Screening a Deal Step by Step — Screen a rental deal with a traceable input sheet, cautious cash-flow maths and a list of facts to check by hand.
- Can AI Draft Rent Arrears Letters That Stay Within the Rules? — A four-stage arrears letter ladder with ledger checks, sample wording and a firm legal-review stop point.
- Can AI Draft a Tenancy Agreement After the Renters' Rights Act? — AI can draft the first version of a periodic tenancy agreement, but it will not know the new rules. A checklist of what to verify before you sign.
- Can Landlords Use ChatGPT for Tenant Notices? The Validity Checks — A landlord's message library plus the validity checks every AI-drafted notice needs before it is sent, and a list of what never goes into the chat.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Stessa pricing page (stessa.com/pricing), checked 4 November 2026; OpenAI and Anthropic help centres for file handling and data settings. Amounts and the three-property portfolio in the worked example are illustrative.