Pavel Samsonov,
product designer

An annotation tool for extracting and structuring data in financial documents

To comply with non-disclosure agreements, all confidential information has been omitted from this case study.
  • A major bank wanted to automate invoice handling using Workfusion. I led a team to close the gap between our existing tool and the client's needs. more
  • User research revealed a set of pain points, and with the help of stakeholder interviews I prioritized the issues that were the most important to solve. more
  • I focused the design on optimizing frequently repeated actions, and presenting detailed information about the document in a more visible way. more
  • The tool that my team created allowed the customer to process invoices 5 times as quickly as before. more

Context

Workfusion was contracted by a major international bank to automate invoice handling within their systems. The bank's existing process required a lot of manual work by junior business analysts, and though this process was outsourced overseas, it was still slow and costly. The client wanted to use Workfusion's automation tools to reduce headcount and improve processing speed.

Stakeholders at Workfusion decided to use an information extraction approach – text from each document would be annotated by a human, and machine learning would learn from this tagging and eventually annotate most documents by itself. Workfusion's existing information extraction tool was originally designed for dividends, which involved a much smaller amount of text and tags. I was asked to design the next version of the tool, which would be able to handle larger, more complex documents and tags without making the user experience more complicated.

My roles

Discovery

I performed domain research to better understand the client's use case: I examined the types of documents that would need to be extracted, and performed a PACT analysis on the potential users of the tool and the contexts in which they would be using it. I also reviewed the competitive landscape to learn about existing tools. To get a better idea of the problem's scope, I conducted stakeholder interviews, bringing together Customer Success team members, developers, product owners, and clients. I designed and ran think-aloud tests with existing users of the tool, to find real-world issues with the application.

After collecting this data, I combined it with priorities gathered from stakeholders using affinity diagramming. This technique helped me consolidate different kinds of data into several general conclusions. I discovered two primary issues. Firstly, when the document was large and complex, the tool did not offer help to the user. A lot of time was taken up by scrolling up and down the page, and repeating the same actions over and over. In addition, these complex documents frequently had problems with OCR – during the conversion from PDF to text, they sometimes became garbled, or the layout of the document broke. As a result, the handling time for each document was unacceptably long, and the client's employees did not like using the tool compared to their existing, familiar solutions – even if those solutions were much slower.

Design Process

As the feature owner, I set two experience principles to guide this project towards the goal of accelerating document extraction speed.

The first experience principle for this project was visibility of information. The original use case for the tool had only a short paragraph of text and a few tags for the user to annotate that paragraph's key data. The client's financial documents often consisted of several pages, multiple tables, dozens of data tags, and the original PDF file. Helping the user access all this information would reduce the amount of time spent on each document.

The second experience principle was efficiency. A lot of use cases for this tool involved users repeating the same actions over and over (for example, tagging each item in the Country column of a table as the name of a country). Incorporating batch actions – whether interacting with multiple elements at once, or using automation techniques within the UI - would drastically speed up document extraction and make it less tedious for the tool's users to perform these tasks.

With these principles in mind, I created sequence diagrams that described the steps users needed to take in order to achieve their goals. Based on these diagrams, I created wireframes and then mockups. I developed prototypes ahead of the development team's progress, to test with users and make necessary corrections to the designs. For features where precise UX interactions were necessary, I collaborated with the developers in writing code. I ran the project on a lean methodology, focusing on stable releases of key features so that I could have regular demo meetings with stakeholders.

Outcomes

When all stakeholders were satisfied with testing results and stability, the information extraction tool was released to the client's internal workforce. Users were able to extract data much faster than before – a document that used to take 30 minutes with their original in-house method could be done in as little as 30 seconds. Ultimately, this project led to an 80% reduction in full-time employees dealing with document extraction.

Because of the usability improvements, both the client's workforce and Workfusion's in-house users were very happy with the new tool. The team received many requests to distribute this version to projects for other clients.