How Does Robotic Process Automation Become Intelligent? Part 2

Pierre Col

Part of the “Intelligent RPA” series about the evolution of robotic process automation and its potential impact on digital transformation.

In a previous blog post, we examined robotic process automation and the difference between attended RPA and unattended RPA. We also discovered how RPA benefits are fostering digital transformation of large organizations.

To explore the future of RPA, we asked Sebastian Schrötel about his vision for making RPA intelligent. As SAP’s head of intelligent robotic process automation, Sebastian explains how RPA could reach an unprecedented level of efficiency.

Sebastian, RPA bots are interacting with business applications via their user interface. Mimicking a human user working on a PC, the RPA bot interprets user interface (UI) components and extracts the data it needs to execute tasks, virtually copying, pasting, clicking, etc. Is it efficient and reliable? And could it be improved?

That UI-based approach works, and it’s the only one to handle legacy applications such as those provided on mainframes or old versions of Windows business apps that have not been updated for years. This approach can be relatively robust if the RPA bot interprets the structure of the application UI by parsing the domain object model (DOM) to extensively capture the logical structure of the screens and the meaning of every UI component to use it according to its purpose.

We consider APIs the best way to interact with the applications whenever they provide APIs. First, an API guarantees upward compatibility when the application is upgraded, which eases bot maintenance and improves their reliability in the long run. Don’t forget that typical RPA bots interact with five to eight different applications. Second, it has lower response time and better performance. Not only does it save resources from your IT infrastructure, but it also lets your bots do more work in the same time.

Let’s talk about data. RPA bots manipulate structured data stored in various databases and application silos. But now they also have to cope with more and more unstructured data, such as images, text, and speech, often coming from mobile devices. What could we imagine here?

There are a lot of business scenarios involving unstructured data in enterprise processes.

With images, you might have to recognize and compare objects or extract data from a document captured in a photo, like a customer ID card during a know-your-customer (KYC) process.

With text, you might need to do automatic classification of documents, or extract keywords and metadata from an invoice or a purchase order, or even identify the tone and the intention behind the document if it comes from an end customer to customer service.

With speech, you could eventually convert speech into text to feed your systems, or the other way around to deliver messages to customers or employees of your organization.

How can we give RPA bots advanced capabilities for serving those business scenarios?

For this kind of specialized cognitive task, the answer is artificial intelligence. AI has benefitted from huge improvements during the last few years. Thanks to the huge amount of data available now to train the models, machine learning and deep learning algorithms have reached a very good level of confidence, beating the human brain in most cases.

And if we try to think further than handling unstructured data, we could imagine that within a few years, AI could help bring self-learning capabilities to RPA bots. Bots could learn by understanding what a human user is doing to replicate some tasks. Bots could even be able to adapt themselves to some minor changes in their working environment – for instance, to handle exceptions or updates in the applications they interact with.

When it comes to robot-human interaction, with the RPA bot being a kind of “digital assistant” for a human employee, could it be better?

I do think so. Instead of interacting with a user via a regular UI, an RPA bot could offer a high-quality service experience by handling a normal conversation with questions and answers. Some chatbot technologies are not based on traditional natural language processing (NLP), which requires specific developments and configuration for every language to handle. Best-in-class chatbots leverage machine learning to be fully language-agnostic. And the use of preconfigured chatbots for specific industry sectors reduces integration time and speeds up deployment.

Then what do you think is the next big step for RPA?

I would say that “RPA + Artificial Intelligence + Conversational AI = Intelligent RPA” is the right equation for 2019.

Thanks, Sebastian!

Intelligent robotic process automation

Can you imagine what your organization could achieve with the powerful combination of RPA bots able to interact with all the applications of your information system, empowered by a full range of AI technologies to manipulate unstructured data, with chatbots to simplify user interactions via any device?

That’s exactly what SAP just introduced. We’ve worked hard to seamlessly integrate our SAP Intelligent Robotic Process Automation solution with existing SAP Leonardo machine learning and SAP Conversational AI.

And the cherry on top – because more than 60% of all RPA bots deployed worldwide interact with SAP systems, SAP also provides prebuilt automation scenarios for SAP products.

This allows your organization to set up and deploy agile teams of “digital assistants” and “digital workers,” leveraging their virtual brain to understand and decide, virtual ears and mouths to communicate, and virtual arms to execute digital tasks.

Those new-generation bots will make your organization even more intelligent and more efficient.

Learn more

For more information, please visit our SAP Intelligent Robotic Process Automation web page.

And please listen to the replay of our “Pathways to the Intelligent Enterprise” Webinar, featuring Phil Carter, chief analyst at IDC, and SAP’s Dan Kearnan and Ginger Gatling.

This article originally appeared on the SAP Analytics blog and is republished by permission.

Follow SAP Intelligent RPA on Twitter, Linkedin, YouTube

Pierre Col

About Pierre Col

Pierre Col is head of communications for Intelligent Robotic Process Automation at SAP, formerly chief marketing officer for Contextor. He has an extensive 30+ years background and expertise in strategic marketing, field marketing, Web marketing, lead generation, corporate and business communications, analyst relations, investor relations for Internet, telecom and IT companies.