Deciphering the Data Problem for RPA

[Anonymous]

In my time at Blue Prism, I have rarely seen as much excitement about a product announcement as when we announced Decipher at Blue Prism World 2019.

Since joining Blue Prism almost 3 years ago, I have had many conversations with our customers and partners and there are a small number of common themes that arise – when exploring the limitations and challenges of RPA. Today, I will focus on the problem of Data acquisition, transformation and understanding.

We all dream of a world where all data is in digital form at source and is at least semi-structured. The reality is very different. Many processes are still “paper based” in some form. These documents are often in highly unpredictable and variable formats, quality and structures. Add to that the free form text from fields in your systems of record, chatbot interactions, emails and a myriad of other interfaces and you have a mess of data that often hampers your ability to transform your processes with RPA.

We only have to look at our Digital Exchange statistics to see how prevalent this problem is – over 60% of downloads from the DX are of Visual Perception skills that relate to acquiring or understanding data. Most of our customers are looking at how they can expand beyond simple, structured and rule-based processes. To support this next evolution, our view is that certain capabilities need to be considered as “table stakes” within an Enterprise RPA platform. The ability to ingest, process and understand data is one of those capabilities. This is why our customers will receive this basic capability included in their license.

Of course, this challenge is too big for us to solve on our own. We have many partners in our Technology ecosystem who have built their software and business on the basis of solving this problem – Abbyy, Moonoia, Rossum, Ephesoft, Hyperscience and Vivado – to name a few. There are also an increasing number of easily accessible and intelligent cloud APIs. You only need to look at the new offerings from the major cloud vendors - such as Microsoft Form Recognizer and AWS Textract - to see the impact that AI is having on this space and the potential for commoditization in this space. That said, this is a complex problem and we believe that there will be difficult to solve problems that will continue to require a more specialized approach – complex or domain specific documents, poor quality images, or handwriting recognition – for example.

So, consistent with our partner first strategy, one of the most important attributes of our Decipher roadmap is that it will be extensible and will enable easier integration with our partners in this space. We call these extensions “Decipher Skills”. These will come in 2 forms. Firstly – the ability for our partners or users of Decipher to train the Decipher ML to classify and extract data from new types of documents. The resulting ML models will be exportable and can be made available via the Digital Exchange. The second will be the ability to tightly integrate 3rd party capabilities directly within the Decipher workflow. By ensuring a consistent approach to data ingestion, output and training, we will streamline and optimize the approach of how your Digital Workforce handles data; even when specialist solutions are required to augment Decipher’s capabilities.

Lastly, whilst it’s true that the first beta of Decipher, due later this summer. will be pre-trained on Invoices and many customers will gain significant value from that alone, the vision for Decipher is much bigger. Through our embedded ML from the Blue Prism AI Labs, business-led user experience and connected-RPA strategy - enabled by Decipher skills - we will enable you to create a recipe for Deciphering the data and language of your business. Ultimately, this will usher in a new wave of Intelligent Automation and will allow you to free up your Digital Entrepreneurs to work on higher value tasks.