AI Automations
AI GL Code Assignment
Topkey uses AI to automatically suggest a GL code for each transaction, pulling from your connected chart of accounts and learning from the coding decisions you make over time. This reduces manual categorization work and helps keep your books consistent across properties.
How GL Code Prediction Works
When a transaction comes in, Topkey analyzes several signals to select the best matching GL code from your chart of accounts:
- Merchant name and location — the business where the transaction occurred
- Transaction description and amount — the details of what was purchased
- Past coding decisions — how you or your team have categorized similar transactions from the same merchant
- Card usage patterns — the typical purpose of the card used for the transaction
- Receipt details — if a receipt is attached, its line items and descriptions are factored in
Topkey also calculates a confidence score for each suggestion, so you can quickly see how certain the AI is about its recommendation.
Master GL Codes
Behind the scenes, Topkey maintains a set of master GL codes — a standardized internal taxonomy that maps across different accounting integrations. When you connect QuickBooks, NetSuite, or a custom chart of accounts, Topkey maps your specific GL codes to this master taxonomy.
This means that if you switch accounting integrations or update your chart of accounts, Topkey's predictions stay consistent. The AI learns from your coding history at the master GL code level, so institutional knowledge isn't lost when integrations change.
Accepting a GL Code Suggestion
When Topkey assigns a GL code automatically, the suggestion appears on the transaction detail view with the recommended code and account name. You have two options:
- Accept the suggestion — click the accept action to confirm the AI's recommendation. The GL code is applied to the transaction and synced to your accounting integration.
- Override the suggestion — select a different GL code from your chart of accounts. The transaction is coded to your chosen code instead.
Accepting suggestions takes just one click and immediately syncs the GL code to your connected accounting system.
Providing Feedback to Improve Predictions
When you choose a different GL code than the one Topkey suggested, Topkey treats your selection as feedback and uses it to improve future predictions for that merchant.
After rejecting a suggestion, a feedback modal appears where you can type a plain-English explanation of why a different code is correct — for example, "This Home Depot purchase is for property maintenance, not capital improvements." This explanation is used to retrain the model and can trigger an immediate updated suggestion for the transaction.
The more consistently you code transactions from a particular merchant, the more accurate Topkey's suggestions become over time.
Regenerating a Suggestion
If your chart of accounts changes, or a suggestion looks incorrect, you can request a fresh prediction without waiting for the next sync cycle:
- Open the transaction detail view.
- Locate the GL code section.
- Select the option to regenerate the suggestion.
Topkey clears the existing suggestion and runs the AI prediction again using the latest version of your chart of accounts and any new context available for the transaction.
Regenerating a suggestion is useful after adding new GL codes to your chart of accounts or after attaching a receipt that provides additional context.
GL Code Assignment for Split Transactions
When a transaction has been split across multiple lines, Topkey generates a separate AI GL code suggestion for each individual split. You can accept, reject, or override the suggestion on each split line independently.
This allows you to accurately categorize a single transaction that covers multiple expense types — for example, a supply run that includes both cleaning products and a capital item.
When you accept an AI suggestion on the parent transaction, Topkey automatically copies that suggestion to any splits that haven't already been coded.
Filtering Transactions with Natural Language
Topkey includes a natural-language transaction filter that works alongside GL code assignment. Instead of configuring filter options manually, you can type a plain-English query — such as "Home Depot purchases over $500 last month" — and Topkey's AI will automatically translate it into the appropriate transaction list filters.
This makes it easy to review a specific group of transactions before bulk-reviewing or updating their GL codes.
Frequently Asked Questions
Why did Topkey suggest a particular GL code for my transaction?
Topkey selects the suggestion based on a combination of factors: the merchant name, transaction amount and description, how similar transactions have been coded in the past, the card used, and any attached receipt. If you want to understand a specific suggestion, you can reject it and use the feedback modal to provide context — this also helps Topkey learn the correct code for future transactions from that merchant.
How do I change the default GL code for a specific merchant?
There is no separate setting for per-merchant defaults. Instead, consistently accepting or overriding suggestions for a merchant trains Topkey's model to predict the correct code automatically. The more consistent your coding history, the more reliably Topkey will suggest the right code for that merchant going forward.
What happens if no matching GL code is found?
If Topkey cannot confidently match a transaction to a GL code in your chart of accounts — for example, because the chart of accounts is missing a relevant category — the transaction will be left without a suggestion. You can still manually assign a GL code from your chart of accounts, and you can also use the regenerate option after updating your chart of accounts to get a fresh prediction.