Table of Contents
I. Overview
AI Expense Type Coding helps users code expenses faster by automatically recommending Expense Types based on historical expense data and learned coding patterns.
The feature is designed to improve coding efficiency while maintaining user control. AI-generated values remain fully editable and are clearly identified throughout the system, allowing users to review, accept, or modify recommendations before submission.
In addition, the system provides transparency through audit history and workflow acknowledgements, ensuring users can easily distinguish between AI-generated and manually entered values.
II. Getting Started
To turn on AI Expense Type Coding for your account, reach out to your ProSpend account manager or support team - they'll get it set up for you.
A few things to know before you start:
- Your data is used only to train your own organisation's AI model. It's never shared with, or used to train, another client's model - your recommendations come from a model built solely on your own expense history.
- You'll need at least 3 months of expense history. The AI needs enough past data to learn your organisation's patterns before it can start making recommendations.
- Initial setup takes a little time to process once requested - we'll let you know when it's ready.
- The model keeps learning automatically. Once it's running, it updates itself on a regular weekly schedule, incorporating new expenses (and any corrections your team makes) so recommendations keep improving over time.
III. High-Level Features
Feature |
Description |
|---|---|
AI Expense Type Recommendations |
Automatically recommends an Expense Type based on learned coding patterns and historical expense data |
Editable AI Recommendations |
Users can accept, modify, or replace AI-generated values before submission |
AI Confidence Indicators |
AI-generated values are visually identified within the application |
Audit History Tracking |
Records whether values were generated by AI or subsequently modified by a user |
Workflow Acknowledgement |
Users are required to acknowledge AI-generated values before progressing workflow actions |
Claim Search & My Spend Indicators |
Displays AI indicators in lists when claims contain AI-generated values |
IV. UI Navigation
1. AI Expense Type Recommendations
AI recommendations are generated for new expenses created through UI, file import, or document scanning.
AI recommendations are only generated when:
- No initial Expense Type has been populated or mapped.
- Expense types defaulted based on the Expense Type Match rules or MCC rules will take precedence over the expense types recommended by ProSpend AI.
- Supplier and Expense Amount fields need to be populated.
The system also uses below data to recommend an Expense Type:
- Entity/Department
- Claimant
- Claim Types:
- Cash Claim
- Card Claim
- Travel Claim
- The Supplier carries the highest weighting in the prediction model, with other expense details used to further refine the recommendation.
- Supplier name matching: ProSpend AI accounts for minor variations in how a supplier's name is entered - capitalisation, punctuation, typos, suffix and prefix additions, and word-order changes - so recommendations stay consistent even when the same supplier appears slightly differently across transactions.
Where the AI recommendations are shown:
a. Claim - Expense Rows
b. Expense Details
AI Recommendation Indicators
The system displays visual AI indicators to show the recommendation confidence level of AI-generated values.
Confidence thresholds are predefined by the AI prediction model and are not configurable at the client level.
Indicator Status |
Description |
|---|---|
|
Analysing
|
The system is currently evaluating the expense and generating recommendations. A recommendation has not yet been determined.
|
|
High Confidence
|
The AI has a high level of confidence in the recommendation and automatically populates the field with the recommended value. AI has a confidence score of 90-100%. |
|
Medium Confidence
|
The AI has moderate confidence in the recommendation and automatically populates the field with the recommended value. AI has a confidence score of 70-89%. |
|
Low Confidence
|
The AI has identified a possible recommendation but has low confidence in its accuracy. The recommended value is displayed to the user but is not automatically populated. Users can choose to accept the recommendation or enter their own value. AI has a confidence score of 1-69%. |
|
Unable to Recommend
|
The AI was unable to determine a suitable recommendation based on the available information and historical data. No value or recommendation is provided. |
Editable AI Recommendations
AI-generated recommendations remain fully editable and do not restrict user input.
Users can:
- Review AI-generated recommendations
- Replace the recommendation with their own selection
The AI acts as a recommendation only and never prevents users from entering or selecting alternative values.
2. Workflow Acknowledgement
Before certain workflow actions can be completed, users must acknowledge AI-generated values.
- Submit
- Approve
- Finalise
The acknowledgement confirms that:
- AI-generated values have been reviewed
- User accepts responsibility for submitted coding
- Claim can proceed to the next workflow stage
3. Audit History Tracking
The system maintains a complete audit trail of AI Coding activities to provide transparency into how expense values were assigned and subsequently updated.
Audit history allows users to understand whether a value originated from AI recommendations or was manually entered or modified by a user.
What is Recorded
- AI-Generated Recommendations: When the system automatically generates an Expense Type recommendation, an audit entry is created (This entry is only visible to admin)
- User Modifications: If a user updates an AI-generated value, a separate audit entry is created to capture the change
- AI Acknowledgement: When a user acknowledges that a claim contains AI-generated values, the acknowledgement is recorded in the audit history.
4. My Spend Page
My Spend page provides visibility into claims that have AI Expense Type Coding. An AI indicator is displayed within the Alert column for claims that contain AI-generated values such as Expense Type recommendations. This allows users to quickly identify expenses that have been coded with AI assistance and may require review.
Filtering AI-Assisted Claims
Users can also locate AI-assisted claims through the Alert Filter dropdown.
