Introduction to MVP creation

The core idea behind an MVP is to develop a product with just enough functionality to address the most critical needs of your target users, allowing you to validate assumptions and learn from real-world usage.

By focusing on essential features, an MVP lets us bring the product to the market quickly, where we can test it, gather feedback, and iterate based on real customer interactions. This approach helps minimize the risk of investing time and resources in unnecessary features while ensuring that the most important aspects of the product are thoroughly validated.

Technical Implementation of the MVP UI for Coding Proposals

In the MVP, we focus on the core feature of providing coding proposals to streamline the customer’s invoice handling process. The aim is to prefill the invoice with coding proposals before the end-user interacts with it. This allows the user to focus primarily on reviewing and adjusting the proposal rather than manually coding from scratch.

Confidence Levels and Visual Cues

Kaunt provides coding proposals with three confidence levels: HIGH, MEDIUM, and LOW. Each of these levels represents how confident the system is in the correctness of the proposed coding. To help users quickly assess which proposals require attention, we recommend color-coding the confidence levels in the following way:

• HIGH confidence: Green (indicates that the proposal is likely correct and may not need user review).
• MEDIUM confidence: Yellow (suggests that the proposal is reasonably accurate but might require manual review or correction).
• LOW confidence: Orange (implies that the proposal is uncertain and should be carefully reviewed by the user).

This approach provides a simple, intuitive way for users to identify potential issues, streamlining their workflow and making it easier to prioritize which dimensions to review.

Implementation of the UI for Coding Proposals

There are multiple ways to present coding proposals in the user interface, depending on the design flexibility of the existing system.

Using Drop-Down Menus

• If the system supports drop-down menus for coding dimensions, we recommend displaying up to three coding proposals per dimension. The proposals can be presented directly in the drop-down, with each option color-coded according to its confidence level.
• This approach allows users to quickly select a different proposal if they believe the prefilling is incorrect. They can see which alternative proposals are available and their associated confidence levels.

No Drop-Down Menu Option

• If the existing system does not allow for drop-down menus, we recommend defaulting to the highest confidence proposal for each dimension. In this case, the highest confidence proposal will be automatically applied to the invoice, and users can manually change the coding if necessary.
• The color-coding system will still apply, with the prefilling color indicating the confidence level of the automatically selected coding proposal.

Marking AI Automation on Posting Lines

• In addition to color-coding the proposed values for the individual dimensions, it is often useful to add a visual indicator on the posting line itself to signify that AI automation has been applied. This could be an icon, label (e.g., "AI Coded"), or any distinct visual marker.
• This marker helps users understand that some level of automation was involved in generating the coding proposal, even before they interact with the individual coding fields. It increases transparency and user confidence by clearly signaling where AI has assisted the process, allowing for a more informed review of the proposals.

Why This Approach?

• Simplicity: The UI design for the MVP is kept simple and intuitive. Users can quickly see what’s prefilled and whether they need to intervene, minimizing confusion or cognitive overload.
• Efficiency: By reducing the need for manual input at the start, this approach speeds up the invoice handling process while still allowing for user control and oversight.
• Focus on Core Functionality: The MVP is designed to focus on delivering value by providing reliable coding proposals. More advanced features like fully automated coding without user interaction can be introduced in later versions, based on feedback and testing.

By structuring the MVP UI in this way, we ensure that end-users have the tools they need to interact with the invoice in a straightforward manner, allowing them to benefit from automation without losing the ability to intervene where needed.

Coding Proposal Outputs

Kaunt provides up to three coding proposals for each invoice, each assigned a confidence level (HIGH, MEDIUM, or LOW). These proposals can be utilized in three distinct ways: Individual Dimensions, Dimension Combination, and Proposal Groups. Each method offers a unique approach to handling invoice coding based on how you want to treat the dimensional relationships and groupings in the coding process.

Individual Dimensions

When using the Individual Dimensions output, each dimension (e.g., GL account, Cost Center, Tax Code) is treated independently. There is no correlation between the dimensions in the output. This means that even if a specific combination of these dimensions has never been used together, Kaunt can still provide proposals for each dimension on its own.

When should I use this output?

• New Combinations: Ideal for cases where your coding practices frequently introduce new combinations of dimension values.
• Handling Poorly Performing Dimensions: Useful for isolating problematic dimensions while automating well-performing ones.

Dimension Combination

In contrast to Individual Dimensions, the Dimension Combination output provides coding proposals that maintain correlation between the dimensions. The output is restricted to combinations of dimension values that have been seen before in previous codings.

When should I use this output?

• Established Combinations: Effective when your coding practices rarely introduce new combinations of dimension values.

Proposal Groups

When dealing with multi-line invoices, it's common for multiple invoice lines to be coded in the same way. Proposal Groups group these lines together based on the Dimension Combination output, producing a single proposal for the group rather than individual proposals for each line.

When should I use this output?

• Grouped Posting: Ideal for grouping similar invoice lines under the same coding scheme.

Feedback Loop

In an MVP, especially during pilot phases with customers, gathering feedback is critical to refining the product and improving its core functionality.

Metrics to Track

• Proposal Precision: How often users override or modify the coding proposals.
• Automation Rate: The percentage of invoice volume that is successfully automated.

Benefits of the Feedback Loop

By implementing a posting feedback loop, Kaunt can:

• Analyze how closely the proposals align with the final postings.
• Identify patterns requiring adjustments.
• Fine-tune the algorithm to improve accuracy.

The feedback loop is essential for continuous improvement, allowing Kaunt to enhance its proposal system based on real-world user interaction and invoice outcomes.