Context
PwC is a global consulting company providing independent audit services. One of the core workflows is the verification of a client’s financial documentation. This process is carried out in several stages and performed by two independent teams to minimize errors caused by the human factor. The work itself is highly repetitive: auditors manually compare values across multiple documents, statements and financial forms.
The company aimed to reduce the time required to complete audit projects and increase employee efficiency through digitization and automation of the document review process. As part of this initiative, the RIM (Risk & Information Management) platform was developed to integrate document management, automated verification, discrepancy detection and quality control into a single environment.
Year
2017 – 2018
Client
Pwc
Challenge
Audit review is time-consuming due to the volume of documents and the requirement for duplicate checks. Auditors spent a significant portion of their time searching for specific values, comparing them and recording outcomes. Errors were not caused by lack of expertise, but by cognitive overload and the repetitive nature of the work.
The goal was to reduce review time while maintaining accuracy and compliance, and to lessen the impact of the human factor on the results.
Approach
Before moving into design, I immersed myself in the internal audit workflow to understand the real working context: how documentation is handled, which systems are used, and what regulatory and procedural constraints define the process. I conducted interviews with employees at different levels to observe how the work is actually performed. Specific document examples and internal data cannot be shown due to NDA, but the core patterns and workflow structure can be described.
We identified a recurring repetitive workflow that consumed a significant amount of time: locate a value in the audit sample, locate the source document, find the corresponding value, hold both values in memory, compare them and record the result. This sequence is repeated many times throughout the day, creating cognitive overload and increasing the risk of mistakes.
Based on this insight, I designed the system to remove cognitive load and eliminate manual comparison steps. The platform organizes documents, communication and verification results within structured projects. Machine learning and computer vision automatically extract and match values, while the interface displays them side by side to support immediate decision-making without memorization. The platform also stores comments, version history and intermediate verification steps, making the audit process transparent, reproducible and suitable for collaborative work.
Impact
The platform significantly reduced completion time for common workflows and decreased the likelihood of human error. Reviewing 10 values on a page decreased from ~45 seconds to ~6 seconds. Document identification time decreased from ~18 seconds to ~2 seconds. Human-factor errors decreased from ~4% to ~0.2%. Overall productivity of audit teams increased by approximately ~15x.
The platform made the review process faster, more accurate and more predictable, while preserving necessary oversight and quality standards.