Next Generation - General explanation on how the AI assisted Bank Reconciliation works

This article aims to provide a general explanation on the new Bank Reconciliation process and links out to user guides which take you through the step-by-step guides to set it up and get you going. In order to fully understand how the Bank Reconciliation process works with AI, we strongly recommend that you read through this article.

Bank Reconciliation is the process of comparing the company’s internal financials records with its bank statement records to ensure records align.

The new process introduces significant improvements to the existing Bank Reconciliation functionality, including a new design and the integration of the AI (Artificial Intelligence) Auto Matching services providing a more intuitive, streamlined and efficient reconciliation workflow.
​
​

The AI engine intelligently matches Bank transactions with Cash Book transactions, based on transaction details such as amounts, dates and descriptions.

It is important to mention that this is an AI assistant service, designed to assist in the process of a bank reconciliation, therefore the user of the service must review the matches produced and provide feedback, accepting, rejecting or correcting (through manual reconciliation) the matches suggested. The AI service relies on user feedback to track and improve the success rate of the matches.

This AI service matches your bank transactions to your cash book transactions. For bank transactions, the following fields are used in the matching rules:

For the Cash Book transactions, the following fields are used in the matching rules:

For a match to be valid, it must as a minimum satisfy the following criteria:

Many automated bank reconciliation systems can only reliably work with one-to-one matches, as most matches are one to one type, and they are faster to be identified. Many to one or many to many match types are often resolved using history from previous successful matches manually performed or based on pattern recognition.

There are three main match types that our AI service can return which are:

Matches are generated using "matchers", internal functions which follow strict conventions on deciding the matches to be suggested for you. Each matcher is executed in sequence and forwards its results to the next matcher, to ensure that matchers only focus on transactions which remain unmatched.

These matchers are broadly categorised into the following types:

When matches are accepted, rejected or corrected (through the manual match), the AI service creates history records that are used to improve the success rate of matches suggested.
Historic matches are occasionally pruned, to ensure that only the most relevant history is used and to replace old history with new patterns. Only history not used for a period of time is pruned while meaningful historical patterns are leveraged for accurate transaction matching.

The AI service learns based on your feedback from previous matches that you have accepted, rejected or corrected through manual matching. Since match history can be vast, only a selection of the most relevant history is used.

It is important to mention that AI services cannot learn immediately after just one manual match correction as repeated actions are needed to define patterns. Also, it is not possible to provide an accurate estimated time frame of how long the AI will take to learn, as statement transaction details can be different between bank accounts, customers or even between months.
In conclusion, the smaller and simpler the data is, the more quickly the AI will learn.

As this AI service is performing a variety of different matchers including the AI ones, it can be slower than traditional bank reconciliation services, depending on the size and complexity of the data used.

To use the matchers described in the table above, the data is uploaded in advance of the match request being run and enriched to be used by all the matchers.

The advantage lies in the accuracy of the matches, reducing human intervention over time. The average time to perform matches, can vary depending on number of transactions and complexity of the matching logic.

The average matching time does increase with the number of transactions, however the actual impact depends on whether the system uses simple or complex matching algorithms.
The AI service can introduce slightly higher processing time in exchange for improved accuracy and flexibility especially for complex and ambiguous data used.

New User Interface Design, New Process Flow and Customisation

We have simplified the navigation, introducing a clearer and more intuitive user interface design, following a tabbed layout to easily switch between tabs and to have a clearer view of the reconciliation progress and actions required.

With one click approval of suggested matches, the reconciliation process is now faster, as all actions required, are performed in a single screen.

You can view the suggested matches in Matched Tab, you can action any discrepancies in the Unmatched tab and the reconciled transaction are displayed for you under the Reconciled Tab.

You can configure the displayed columns the same way as you can do so for enquiries, processing and record list screens using the existing Configure Column functionality.

The quick search options, filtering and sorting capabilities can now be used through the bank reconciliation process, allowing you to easily identify the needed transactions, especially during the manual matching step.
You can access the Auto Match option throughout the whole reconciliation process and you can action it any time to reconcile new Cash Book account transactions posted, or whenever there is need to auto match transactions.

Increased Efficiency:

Higher Accuracy:

Time Saving: