Audit Intelligence Engine

## Inspiration Traditional audit tools generate large volumes of flags, reports, and dashboards, but interpreting those outputs still requires significant manual effort. The idea behind Audit Intelligence Engine was to explore whether audit outputs could become more decision-oriented instead of just data-heavy. The goal was to build a system that not only identifies unusual transactions, but also explains why they matter

## What it does The application: detects unusual transaction patterns identifies missing documentation and approval gaps highlights high-risk vendors and transaction concentrations prioritizes areas requiring review generates audit-style reasoning and summaries visualizes risk trends over time through a risk timeline

## How we built it I used MeDo to build an audit-focused prototype that helps interpret transaction data more effectively. The idea behind the project was to reduce the manual effort involved in reviewing large volumes of transactions and identifying where attention is actually needed. The app analyzes uploaded transaction data and highlights: key anomalies their potential impact priority areas for review It also includes a risk timeline that visualizes how audit risk changes over time and highlights spikes in activity.

To build the project, I structured conversations with MeDo iteratively. I first focused on defining the workflow and analysis logic, then refined the outputs, reasoning structure, and interface step by step. I focused on creating outputs that felt more decision-oriented and easier to interpret. One of the most impressive features MeDo helped generate was the audit judgment summary. Rather than simply flagging transactions, the system generates structured explanations describing why certain patterns may indicate elevated risk and what should be reviewed first. The entire prototype, including the UI, logic flow, analysis outputs, and risk timeline, was built end-to-end using MeDo.

## Challenges we ran into One of the biggest challenges was balancing simplicity with meaningful audit logic. Initially, the outputs felt too generic and overly rule-based. A major focus became improving how the system explains risk rather than simply listing flagged transactions. Another challenge was avoiding excessive dashboard complexity. The interface had to remain clean and interpretable while still presenting enough audit detail to be useful.

## Accomplishments that we're proud of We are particularly proud of: building the entire prototype end-to-end using MeDo creating decision-oriented audit outputs instead of basic anomaly lists generating structured audit judgment summaries with explanations

## What we learned We also learned the importance of iterative prompting and structured workflow design when building applications with MeDo.

## What's next for Audit Intelligence Engine automated invoice/document verification conversational audit querying live collaborative review workflow

Built With

• medo