TDWI | Training & Research | Business Intelligence, Analytics, Big Data, Data Warehousing

Is Robotic Process Automation Intelligent?

Are robotic process automation and artificial intelligence the same? Not yet, but with advances in generative AI, these two technologies are coming closer together than ever before.

• By Troy Hiltbrand
• May 8, 2023

When people think of robots, they intuitively think of artificial intelligence (AI). This is because we have been conditioned by popular culture such as Terminator and Star Wars, where robots are almost sentient. With this frame of reference, people are often confused by the concept of robotic process automation (RPA). They assume that implementation of this technology will bring artificial intelligence to their processes.

For Further Reading: Three Key AI Trends to Watch in 2023 Three Models Leading the Neural Network Revolution How Robotics Process Automation Eases Data Management

Until now, the concept of RPA has been heavily focused on process automation. Its selling point has been around automating routine, short-running tasks where there is not a direct API available for direct system-to-system integration or where there is not sufficient development capacity to connect to the direct API. Much of their work revolves around interacting with software applications, moving data, performing basic actions, such as arithmetic calculations or completing missing data, and responding to business triggers. RPA platforms have allowed businesses to execute these rules-based business processes quickly and with solid levels of reliability and precision. The implementation of these rules-based automated processes is done through the deployment of software robots, or “bots.”

With recent advances in AI, RPA is starting to change. Some of the largest RPA vendors, such as Microsoft and Automation Anywhere, are building steps into their flows that consume AI services to provide business processes with the next level of intelligent capabilities. Let’s look at three leading use cases where RPA and AI are being merged to produce some impressive business outcomes.

Natural Language Generation

With the rapid advancements of large language models (LLMs) such as OpenAI’s GPT and Google’s Bard, the concept of natural language generation (NLG) is garnering significant attention across many industries. By leveraging high-powered machine learning models, narrative responses to text or voice prompts can be generated that appear almost indistinguishable from those of humans. As these services are built with an exposed API layer, RPA vendors are able to integrate these capabilities directly into a workflow. This can be leveraged to generate very human-like correspondence and communication as a step in a process. Whether this includes generating an email, a letter, a text message, or even a voice message that includes many personalized details, generative AI in the form of NLG can create an output that is ready to send.

As a component of the business intelligence process, these communications can also pull quantitative and qualitative outputs from the analytics process into their narratives and prepare them to be delivered in a highly consumable fashion by relevant stakeholders. This has the potential of augmenting the data synthesis and analysis phase of business intelligence and get insightsquicker.

Computer Vision and Natural Language Understanding

One of the other areas where RPA bots have been deployed to save time and improve efficiencies is in the extraction of data. Traditionally, much of this data comes through sources such as email, PDF forms, and Excel spreadsheets. These could be originating outside the company -- such as through vendors or customers -- or inside the company. The unstructured nature of these data sources and their variability make it challenging to create a high level of reliability in parsing the data and preparing it for further processing. With advances in computer vision and natural language understanding, these same LLMs can be leveraged to parse data in its highly variable states and return it cleaned and ready for processing.

By reducing the need for human intervention in the inbound consumption of multiformatted data, the process of analysis and action can complete more quickly. This can shorten the time to make decisions and complete business processes.

Code Generation

Some processes require integration with other systems. In the past, RPA solutions have catered to business users with varying levels of technical acumen. However, even if the systems had an underlying API, these business users were not inclined to start developing low-level code to interact with it. As a result, these RPA solutions were often paired with low-code or no-code solutions to perform some direct system integrations. However, their capabilities have had limits as far as the logic and functionality that could be built. When business users ran into these limits, they would have to engage developers in the IT department, which often introduced delays.

As part of the advancements of LLMs, not only can they generate narrative text, they can also create functioning code. When deployed inside of an RPA, contextually specific code can be generated (in languages such as Java, .Net, or Go) and executed by business users. As this automatically generated code becomes more complex and comprehensive and less error-prone, business users will be able to connect more easily with a wider variety of systems than ever before without the need for heavy involvement of IT. Depending on the nature of the business process, these business users can perform initial code generation and work with developers and the software quality assurance staff to validate and verify its functionality.

Final Thoughts

RPA has been somewhat of a misnomer in the past because it really didn’t leverage artificial intelligence, but this is changing rapidly. RPA vendors are actively integrating LLMs to make processes more powerful than ever before. We might not be at the point where these RPA bots think autonomously, but we are getting closer to being able to deploy more artificial intelligence across the organization to speed up business processes and improve decision-making.