How to Auto-Categorise Revolut Transactions with AI for UK Bookkeeping

The UK Bookkeeper's Dilemma: Untangling Revolut Transactions

If you're running a small business or working as a freelancer in the UK, chances are you've dipped your toes into Revolut. It's fantastic for its ease of use, competitive exchange rates, and multi-currency accounts – a genuine boon for anyone dealing internationally or just wanting a slick banking experience. But let's be honest, come tax time or even just monthly reconciliations, those Revolut statements can feel like a financial spaghetti junction.

Manually categorising every single transaction, especially if you're making lots of small purchases or dealing with different currencies, is a monumental time sink. You're trying to figure out if that "London Coffee Co" debit card payment was for client entertainment, a team meeting, or just your morning caffeine fix. Multiply that by hundreds of transactions, and you've got yourself a proper headache. This is where the promise of AI steps in, offering a genuinely practical way to bring order to your Revolut account and make your bookkeeping significantly less painful.

Why Automate Revolut Categorisation?

For UK small businesses and freelancers, accurate bookkeeping isn't just a nice-to-have; it's a legal requirement and crucial for making informed financial decisions. HMRC wants a clear picture, and you want to know exactly where your money is going. Revolut, with its flexibility, can sometimes complicate this:

• Multi-Currency Mayhem: You might pay a supplier in Euros, receive payment in USD, and cover a subscription in SEK, all from the same account. Tracking the original currency, the converted GBP amount, and the exchange rate manually is a chore.
• Vague Merchant Names: Sometimes the transaction description is clear, sometimes it's an obscure string of characters that tells you nothing about the actual vendor or service.
• High Transaction Volume: Modern businesses often have many smaller, frequent transactions, from software subscriptions to online ad spend. Each one needs a home.
• Time is Money: Every hour you spend manually sorting transactions is an hour you're not spending on growing your business or enjoying your downtime.

This is why AI financial admin has become such a hot topic. It’s not about replacing you entirely; it's about giving you a very smart assistant that can handle the repetitive, rules-based work, freeing you up for the more strategic financial tasks.

Understanding AI for UK Transaction Categorisation

When we talk about using AI to auto-categorise Revolut transactions, we're not talking about a magical black box. At its core, it’s about setting up intelligent rules and patterns that an AI system can learn from and apply. Think of it as teaching a very diligent intern your bookkeeping logic, but supercharged. Instead of you telling it "every time you see 'Tesco', put it as 'Groceries'," you're giving it broader instructions or examples, and it figures out the nuances.

For UK businesses, the goal is often to align these categories with common accounting codes or your specific chart of accounts (e.g., Office Supplies, Travel, Client Entertainment, Software Subscriptions, Bank Charges). This ensures your bookkeeping is ready for your accountant or HMRC, making your HMRC-ready AI expense tracking much smoother.

Getting Your Revolut Data Ready

Before any AI can do its work, you need data. Revolut makes this relatively straightforward.

1. Export Your Statements:

Log into your Revolut business account (or personal account, if that's what you're using for your business finances – though I strongly advise keeping them separate for clarity!). Navigate to 'Statements' or 'Activity' and look for the export option. You'll typically be able to export in CSV (Comma Separated Values) format, which is ideal for spreadsheets, or sometimes PDF. Always go for CSV if available.

When exporting, pick a date range that makes sense for your accounting cycle – monthly, quarterly, or annually. I usually recommend monthly to keep things manageable.

2. Basic Data Cleaning (If Needed):

Once you have your CSV, open it in a spreadsheet program like Google Sheets or Excel. You might find a few columns that aren't strictly necessary for categorisation, like transaction IDs or specific card details. You can hide or delete these to make your working sheet cleaner. The crucial columns you'll want are:

• Date: Obvious, but essential.
• Description/Counterparty: This is the key field for your AI to analyse.
• Amount: The transaction value.
• Currency: Crucial for multi-currency transactions.
• Type: (e.g., Card Payment, Transfer, Exchange) – helpful context.

Building Your AI-Powered Categorisation Workflow (The Google Sheets Approach)

For many UK small businesses, a dedicated accounting software package might feel like overkill initially, or perhaps you're just not ready for the subscription cost. This is where a robust Google Sheets setup, enhanced with AI, can be incredibly powerful. It's a fantastic stepping stone for automate Revolut workflows without a huge initial investment.

Step 1: Define Your Bookkeeping Categories

Before you can categorise, you need a clear list of what you're categorising *into*. These should align with your business needs and HMRC's expectations. Create a separate tab in your Google Sheet named "Categories" and list them out. I often add a column for a brief description of what falls under each, just for my own memory.

• Software Subscriptions: Adobe, Microsoft 365, Zoom, cloud storage.
• Office Supplies: Stationery, printer ink, small equipment.
• Utilities: Electricity, internet, gas (if home office relevant).
• Travel & Subsistence: Train tickets, fuel, hotels, meals during business trips.
• Client Entertainment: Specific meals or events with clients.
• Professional Fees: Accountant, legal advice, consultants.
• Marketing & Advertising: Google Ads, social media ads, graphic design.
• Bank Charges: Revolut fees, international transfer costs.
• Owner Drawings/Salaries: Money taken out for personal use.
• Income: Payments received for services/products.
• Uncategorised/Review: A vital 'catch-all' for anything the AI isn't sure about.

Step 2: Crafting Your AI-Assisted Categorisation Rules

This is where the real magic happens. You'll use a combination of spreadsheet functions and AI to create smart rules. Open your main Revolut transactions tab. Create a new column, perhaps named "Suggested Category."

We're going to build a formula that looks at the transaction description and assigns a category. This can get complex, so we'll break it down.

A. Starting with IF/SEARCH Formulas:

You can start with simple `IF` and `SEARCH` (or `FIND`) formulas in Google Sheets. This is a foundational step before bringing in generative AI to enhance them. For example:

  =IF(SEARCH("TESCO", A2, 1), "Groceries", "")  

This checks if "TESCO" is in cell A2 (your description). If it is, it returns "Groceries." The problem is, you'll have many categories, leading to nested `IF` statements that are a nightmare to write and maintain.

B. Using a Lookup Table with REGEXMATCH (More Robust):

A much better approach is to create a second tab, let's call it "Rules," with two columns: "Keyword Pattern" and "Category."

Keyword Pattern	Category
(?i)tesco|sainsbury|asda	Groceries
(?i)amazon|amzn	Office Supplies / General Purchases
(?i)adobe|zoom|microsoft|slack	Software Subscriptions
(?i)uber|trainline|easyjet	Travel & Transport
(?i)starbucks|costa|pret	Client Entertainment / Refreshments
(?i)paypal|stripe|square	Payment Processing Fees

The `(?i)` makes the match case-insensitive. The `|` acts as an "OR".

Then, you can use a formula like this in your main transactions tab (let's assume your description is in column B and your "Rules" tab is named 'Rules' with patterns in A and categories in B):

  =ARRAYFORMULA( IF(B2:B<>"", IFERROR( VLOOKUP( TRUE, ARRAYFORMULA(REGEXMATCH(B2:B, Rules!A:A)), MATCH(TRUE, ARRAYFORMULA(REGEXMATCH(B2:B, Rules!A:A)), 0) + 1, FALSE ), "Uncategorised" ), "") )  

Okay, that's complex, I know! It uses `ARRAYFORMULA` to apply to a whole column, `REGEXMATCH` to check for multiple patterns, and `VLOOKUP` to fetch the category. If it doesn't find a match, it defaults to "Uncategorised" – which is crucial for review.

C. AI to Generate & Refine Rules:

This is where modern AI truly shines. Instead of manually brainstorming every keyword, you can feed an AI model like ChatGPT, Claude, or Gemini examples of your transaction descriptions and your desired categories.

Here's a prompt you could use (this leans into our essential AI prompts for UK small business bookkeeping):

"I have a list of Revolut transaction descriptions and want to create a set of regular expression patterns for Google Sheets `REGEXMATCH` function. These patterns should match common merchant names or keywords and suggest a category. Please generate these for me based on the examples below, aiming for broad, case-insensitive matches where appropriate. Also, suggest common UK business bookkeeping categories if my examples are limited. Output as a two-column table: 'Keyword Pattern' and 'Category'."

Example Transaction Descriptions: * TESCO STORES LONDON * AMAZON MARKETPLACE UK * ADOBE CREATIVE CLOUD * UBER TRIP LONDON * PAYPAL *STEPHEN SMITH LTD * COSTA COFFEE MANCHESTER * HMRC PAYE

The AI will then suggest patterns like the ones in my "Rules" table above. You can iterate, feeding it more examples or asking it to refine patterns that are too broad or too narrow. This dramatically speeds up the creation of your rule set.

Step 3: Handling Multi-Currency Transactions with AI

This is one of the biggest headaches Revolut users face, and AI can certainly help. Your Revolut export should include the original currency and the converted GBP amount. If it doesn't, or if you need to calculate the GBP equivalent at the spot rate, this is an area where a bit more manual work might be needed, or integration with a tool like Zapier or Make.com to pull exchange rates from an API.

However, for categorisation, the AI focuses on the *description* regardless of currency. A transaction from "Amazon US" for software would still be categorised as "Software Subscriptions," even if it was in USD. The AI doesn't need to perform the currency conversion itself; it just needs to identify what the transaction *is*.

When reviewing, you'll still need to ensure the correct GBP value is recorded for HMRC purposes, but the AI has done the heavy lifting of figuring out *what* it was for.

Step 4: Dealing with Card Payments and Merchant Data

Revolut's descriptions for card payments can sometimes be more informative than direct bank transfers. They often include a merchant name. Your AI rules should specifically target these names. For example, instead of just looking for "COFFEE", target "STARBUCKS" or "COSTA". The more specific your patterns, the better the accuracy.

I've found that sometimes, just searching for the first few words of a common merchant is enough. For instance, "Pret A Manger" might just appear as "PRET AMANGER" or "PRET LONDON". A rule like `(?i)pret` usually catches it.

Beyond the Spreadsheet: Advanced Automation & Review

While the Google Sheets method is robust, you can take it further.

Integrations: Tools like Zapier or Make.com (formerly Integromat) can connect Revolut (via its API or webhooks, if available for your account type) to your Google Sheet or even directly to accounting software like Xero or QuickBooks. This means new transactions could be automatically pulled into your sheet, run through your categorisation rules, and even flagged for review without you lifting a finger. This is getting into proper revolut workflows optimisation.

Regular Review: Even with the best AI rules, you'll always have "Uncategorised" transactions. Set aside time weekly or monthly to review these. This is not a failure of the AI; it's an opportunity to teach it. As you manually categorise these, add the new patterns/keywords to your "Rules" tab. Over time, your system becomes incredibly accurate and tailored to your business.

AI for Anomaly Detection: As your dataset grows, you could even ask an AI model to highlight unusual transactions or patterns. For example, "Look at my 'Office Supplies' category for the last quarter and tell me if any transaction amounts are significantly higher than the average, or if a merchant appears that usually doesn't." This helps catch errors or potential fraud.

Maintaining Accuracy and HMRC Readiness

The beauty of this AI-driven approach for your UK small business bookkeeping is that it creates a consistent and auditable trail. Every categorisation decision is based on a defined rule. When HMRC comes knocking, you can show them exactly how your transactions were categorised.

Remember to keep your "Rules" tab up-to-date. Businesses evolve, suppliers change, and new services emerge. A quick review of your uncategorised items and a refinement of your rules every quarter will keep your system sharp.

Automating this aspect of your financial administration isn't about laziness; it's about working smarter. It allows you to spend less time on repetitive data entry and more time analysing your business's financial health, identifying trends, and making strategic decisions. It's truly empowering when your Revolut statement, once a source of dread, becomes a clean, categorised summary in minutes.