In this article you will learn about the Intelligent Matching functionality including: 

• Overview
• Trigger AI Intelligent match
• Eligible & Ineligible Criteria
• How Intelligent matching works
• Understanding Confidence Score
• FAQ - Intelligent Matching

Overview

This functionality is currently in open Beta phase. 

The Intelligent Matching is an advanced transaction reconciliation feature that uses Artificial Intelligence (AI) and historical learnt data to automatically locate and suggest pairing of a Cash and Business Events transactions on the Match Transactions screen. 

It is triggered via the AI Intelligent button in the Match Transaction screen.

Trigger AI Intelligent match

Navigate to Fund Level > Match Transactions > AI Intelligent

1. Click the AI Intelligent button
2. All suggested cash to Business matches found by Class AI will then be displayed, along with the confidence score, in the Intelligent matches screen for review.

3. Click on Confidence level drop filter to show suggested matches within the confidence level groupings of High, Medium, or Low. 

    

4. Select All, or individual suggested transaction pairs you wish to save.
5. If you wish to proceed with the selected suggested matches then click Save AI Matches. If you do not wish to proceed, click Cancel.

Intelligent Adjustment: Class Intelligence will automatically modify the payment date and/or net payment amount of the business event to the paired cash date and/or cash amount if the AI suggested matches are saved. 

Eligible & Ineligible Criteria

AI Intelligent match eligible criteria:

• 1 Cash to 1 Business Events (e.g. interest to cash transaction) with:
  • Transactions variance in amount of up to $5
  • Within 21 days of the Cash and Business event
  • Opposite directions
  • Partial matching descriptions
• 1 Cash Transaction to 1 Cash Transaction (e.g. transfer between two bank accounts) with:
  • Two bank accounts; and
  • Exact AUD amount; and
  • Same date; and
  • Opposite directions
• Business Events to another Business Events  (e.g. personal contribution against fund expense) with:
  • Two different Event types; and
  • Exact AUD amount; and
  • Opposite directions
  • Within 21 days of each other

AI Intelligent matches will not suggest pairing of:

• Business Events that have a status of Incomplete/pending

How Intelligent matches works

Intelligent matching, unlike traditional static rule-based logic, leverages machine learning models trained on previously confirmed matched, and unmatched transactions, to identify patterns and relationships across multiple data points.

Using past data and the following data points, it estimates how likely the pair is a true match, and assigns a confidence score (from 0 to 1):

• Net Cash and Business transaction amount
• Cash and Business events descriptions
• Cash and Business event dates
• Business event types
• The transactions directions of both Cash and Business events

Understanding Confidence Score

The confidence score represents how strongly the Intelligent Matching model believes that two transactions belong together.

It is calculated by comparing the characteristics of unmatched transactions, on the Match Transactions screen, with patterns the model has learned from previously confirmed matches. These characteristics include factors like transaction amount, direction, timing, and type.

The score ranges from 0 to 1 (or 0% to 100%). The more similar they are to what the model has seen before, the higher the score. The higher the score, the more confident the model is that the two transactions belong together.

FAQ - Intelligent Matching 

Is Intelligent Matching replacing Automatch?

Intelligent Matching is an alternative to using the Automatch button on the Match Transactions screen.

Automatch uses fixed rules-based on date and amount. Intelligent Matching, on the other hand, learns from historical transaction patterns and descriptions to suggest matches more flexibly.
It doesn’t automatically save the matched transactions; those matched by data feeds using Automatch will continue as usual.

Is my data safe?

Our AI training process follows a strict data protection pipeline that ensures no Personally Identifiable Information (PII) or sensitive data is used in model training or stored in the model.

Key Privacy Protections

• No Reverse Engineering Possible: Numerical vectors cannot be converted back to original text.
• No Individual Identification: Mathematical features represent patterns, not specific transactions.
• Aggregated Learning: Model learns from statistical patterns across anonymised records.

How frequently is the model updated?

Our Intelligent Matching model is refreshed every quarter to improve accuracy and performance. When this happens, it may use anonymised data from previously matched transactions to help the system learn patterns in numerical similarity.

This means your transaction data, fully anonymised and securely handled, could contribute to making the model smarter and more effective for everyone.