Cliff Notes for Audit Reports: An AI-Driven Approach
When I was a federal employee, I was involved in my fair share of audits. While they are necessary to provide oversight and to ensure that contractors were being good stewards of resources while meeting the government's requirements, it was a time consuming and tedious task at times. In particular, there was a ton of pre-read documentation that was always supplied during the data call phase before the audit kickoff. You essentially got buried in an avalanche of documentation that you needed to digest to ground yourself on what has been done in the past and what evidence exists as you develop your lines of inquiry and audit/assessment plan.
On more than one occasion, I found myself wishing there was a Cliff Notes version that could roll this up into what I really needed to know. While this seemed like a SciFi pipe dream just a few years ago, it isn't anymore. What I thought was Mission Impossible to me was just another Wednesday for one of our top data scientists, Juliette Easley. Through the naivete of youth and just raw intelligence (the actual data science formula is: Easley > Howerton), she was able to deliver a Proof of Concept (POC) for AI-driven approaches to reading, processing, and parsing PDF audit reports.
The goal of our POC was to accomplish the following:
- Parse out the list of audit findings into structured JSON data
- Conduct a sentiment analysis to measure the positivity and negativity of both the overall document and then each section of the audit report
- Develop visualizations that assist in understanding the sentiment analysis
The results exceeded our expectations and we were able to achieve the following:
- We were able to generate an AI-driven executive summary showing the most positive and negative components of the report
- We were able to parse out the audit findings and turn them into structured data that could then be uploaded in bulk to our automated compliance management platform (Atlasity)
- We were able to build an audit report data model to understand the most positive and negative aspects of any given audit report
You can see POC in action in the following YouTube video:
While the POC showed some promising results, there is so much more we can do with it. Our next steps for a production customer deployment would include:
- Tuning dictionaries to more accurately reflect customer semantics to fine tune the sentiment analysis; every customer has their own lexicon and this could be tuned more precisely to measure sentiment
- Applying batch processing and bulk uploads to migrate large volumes of data into Atlasity. This approach could potentially load years of data via automation, a huge time and cost savings for onboarding a new audit management platform
At the end of this POC, we were able to demonstrate a Cliff Notes solution for using AI to read and process audit reports. In the famous words of Arnold Schwarzenegger on the Simpsons:
While we were excited to make Arnold's dreams come true, we are even more excited to see what we can do with our Atlasity customers as we re-imagine a better way to deliver compliance processes using next generation technology. Less documentation to read with less time and money to load legacy data are game changers for organizations considering a replacement for their existing tool or who may be considering launching their first automated tool to move away from legacy manual processes.
Want to learn more about what C2 Labs and our Atlasity Compliance Management Platform can do for you? Contact us today to learn more or download a free Community Edition to get started on your compliance modernization journey.