Robotic process automation has become a focus point in the global automation landscape. During 2018, RPA was the fastest-growing enterprise software segment internationally, according to research conducted by Gartner; with total RPA market revenue tipped to reach $1.3 billion in 2019.[1]
This stellar performance is rooted in the many business benefits that RPA offers. However, as companies progress further along their RPA paths, many have come to realise that RPA robots – in their pure, unassisted form – can only work with data that is structured.
What is structured data?
This comprises information that an RPA bot can access and process in a predictable and clearly defined fashion. Structured data is organised into a precise format and arranged in a predefined template – often housed within the fields of a traditional database.
If an RPA configuration has access to specific information in a specific place, it can rapidly reach into the right field, retrieve the necessary data and transfer this into another system (or multiple other systems if necessary). These could be accounting packages, time management platforms and even legacy systems without application programming interfaces (APIs) – because RPA solutions work at the same user interfaces that humans do to access data for day-to-day business processes.
For these reasons, provided all data is structured, RPA provides businesses with the tools to automate any routine data-centric task that relies on almost any system – with speed, accuracy and consistency. This not only elevates operational efficiency, but also improves employee performance and satisfaction, because people are no longer wasting their working hours on menial data entry tasks. At the same time, automation reduces the risk of data entry errors and process delays.
A conventional RPA solution is ideal for transactions that require high levels of data accuracy and speed. A bot could, for example, help the accounts department to generate standard invoices in multiple currencies. This type of process could involve a bot logging into an accounting system to retrieve an invoice amount, then logging into a web service like exe.com to request a currency conversion, then transferring the converted currency amount back to the accounting system, where the bot generates an invoice. This entire process can happen in the background without the need for human intervention.
In this case, RPA works well because the process involves structured data and very specific instructions. But what happens when some of the data is not structured?
Managing unstructured data
One common criticism of RPA is the fact that it can’t directly manage the huge amount of semi-structured or unstructured data that businesses deal with on a daily basis.
Documents and content come into organisations in a variety of formats which do not conform to set parameters. Some examples include instructions and queries typed into the body of an email or housed within email attachments such as Word, Excel or PDF files. Departments receive plenty of scanned documents too. Even though all of that data is arriving in a digital format, it is not structured in a bot-friendly way. In other words: an RPA solution alone would not be able to decipher and automatically process that data.
Take invoices as an example. While these typically contain the same type of information – such as dates, vendor details, invoice amounts and purchase order numbers – this content is ordered randomly, according to the vendor’s own system. This means that the RPA bot can’t be programmed where to look for a specific piece of data, unless a template is created for each invoice format, which is a massive task!
One way to get around this is for humans to intervene; and extract and structure the data before handing this over to the RPA bot. Unfortunately, this can be a time-consuming approach, given the high volumes of unstructured documents, content and data that flow into most organisations today. There’s also the added risk of humans making data entry errors. If an RPA bot receives incorrect information, the outcome of the automated process can never be accurate.
Does this mean that RPA use cases are limited? Fortunately not.
A smarter way to get around these data obstacles is to combine RPA technology with an AI-enabled solution. This provides businesses with the intelligent capabilities required to turn unstructured content into a structured dataset before the RPA-driven process begins.
Thanks to the increasing availability and affordability of machine learning, natural language processing and optical character recognition solutions (all different categories of AI), companies can now access the tools they need to sort, decipher and process unstructured data.
Also, because RPA platforms integrate so well with other technologies, AI-enabled tools can be easily embedded into the process automation environment. This means that once the unstructured data has been transformed into structured data using AI capabilities, the standardised dataset can then be passed on to an RPA bot automatically.
This extension of the automation lifecycle is called intelligent automation.
Advantages of intelligent automation
The powerful combination of AI and RPA allows organisations to process data from virtually any source across virtually any system. This includes information that is housed within multiple types of files and documents, including emails, PDF files, scans and photos.
By augmenting robots with cognitive abilities, companies are often able to achieve end-to-end process automation. In other words, they can automate tasks that previously required human input. This means that knowledge workers are able to focus their time on projects that benefit from their unique skills and expertise. The result? The organisation is able to increase productivity and performance across the board.
Below, we have outlined a couple of examples of how intelligent automation can transform common business processes.
Use case example 1: Change of address details
An admin team could have a shared inbox that collects emails from customers requesting a change of their address details. Often, customers communicate their request in the body of the email, using normal conversational language. Alternatively, they may attach an Excel spreadsheet or Word document with the information included in a form or table.
While RPA on its own could not programmatically do anything meaningful with this data, an intelligent automation solution could incorporate a machine learning model that has been trained based on a specific unstructured data set. For example, this model could ‘learn’ how to pick out postcodes from an unstructured spreadsheet, even if these are not in the correct format, contain spaces or dashes, or are in a random location.
Alternatively, this system could also be embedded with an optical character recognition (OCR) solution. This type of AI can recognise and extract data, even when this is embedded in a scanned document. (Some solutions on the market can capture scans of physical documents too, to digitise these before triggering the OCR process.)
The machine learning model could then be trained to take the extracted information, transform this into a structured format and transfer that data to a central repository. From here, the RPA bot could pick up the relevant data from predefined fields and update customers’ address details accordingly in a CRM – or similar – system.
Use case example 2: New account registrations
If an organisation such as a bank or retail group regularly uses digital forms to collect the information required to vet, approve and register customers for new accounts, it’s possible to manage this data-centric process more intelligently.
For example, the company could create a form with predefined fields for capturing customers’ personal information; and then include a feature which allows customers to upload a copy of their passports for a facial recognition and identity verification step.
While an RPA bot on its own can easily pick up the data that is structured in the form fields and engage the banking system to provision this account, it is unable to directly manage the passport screening process. In this case, a machine learning algorithm could be trained how to pick up a copy of the passport photo and compare this with another photo of the individual in order to verify the customer’s identity. If the application passes this verification step, the system could hand the account on to an RPA bot, which could then update the relevant backend system.
Watch a short demo of this solution which adopts RPA and workflow to extract OCR Licence and Microsoft Azure Facial Recognition data to quickly carry out automated due dilligence checks on behalf of the lender.
In closing
RPA on its own offers exceptional value by automating and de-risking standard tasks. However, organisations can substantially amplify this value by combining RPA with AI, in order to expand the amount of data that can be processed in the automation environment – and therefore the variety of assignments that can be handed over to the digital workforce.
As more organisations realise the value of moving from conventional automation to intelligent automation, they will be able to remove many of the obstacles that have been tripping them up on the path to true digital transformation.
Look out for the second article in this series, which will expand on the power of intelligent automation.
Discover Process Automation Solutions
Velocity IT deliver process automation solutions to support organisations on their journey to digital transformation, check out the automation solutions that we have delivered for our clients.