Is RPA the same as AI? What’s the Difference, and What Are the Use Cases?

Data Basics, Use Cases & Projects, Featured Mehdi Hamoumi

The short answer is: No, robotic process automation (RPA) and artificial intelligence (AI) are not the same thing. However, they are two complementary technologies that can increase efficiency in a wide range of business processes.

If you are currently wondering which one your team needs to boost productivity, reduce costs, or unlock previously unreachable use cases, the following analogy might help: AI is the brain; it aims at mimicking and eventually surpassing human reasoning. RPA is the brawn; it enables automation of human actions.

In other words, if you need to automate simple and repetitive tasks, while bringing high accuracy, reliability and traceability, RPA is your choice. On the other hand, if you need to perform more complex reasoning tasks or extract patterns from unstructured data, then go for AI. Of course, do not hesitate to use both if your business problem requires it.

In this article, we will go through a few use cases.  Hopefully by the end of it, you will be better equipped to decide if you need AI, RPA, or both.

robot harm with human-looking hand

RPA: Automating Simple & Repetitive Tasks

RPA uses software robots to automate human actions in business processes that involve interaction with digital systems. These actions are usually simple and repetitive, which makes them prone to human error and can provoke a loss of employees’ motivation and efficiency.

Software robots and RPA on the other hand bring notable benefits: accuracy (by minimizing human error), reliability (by being always available and by reducing delay), traceability (by providing audit trails and logs), and productivity (by increasing processing speed). A few examples of use cases are automating orders, processing payroll, customer onboarding, data validation, etc.

An example of a company providing an RPA software platform is UiPath. Their software integrates both ways with Dataiku DSS, which can bring together the worlds of AI and RPA. More precisely, you can query a Dataiku DSS API node to leverage machine learning models in the UiPath workflows, and import UiPath logs into Dataiku DSS as well as start UiPath Orchestrator processes directly from Dataiku DSS.

→ Learn more about the UiPath Orchestrator Dataiku Plugin

Example Use Case: Insurance Claims Processing

Insurance claims processing is crucial to the insurance business, and it is also a typical process that can be automated with RPA. To understand why, let us go through all the steps that a human would perform for a single insurance claim:

  1. Receive the claim as a PDF in an email.
  2. Download the PDF to a shared drive in a queue folder.
  3. Open the oldest claim in the folder.
  4. Compare it to the company’s data in its internal software.
  5. If no discrepancies are found, copy the claim data into the internal claims software and attach the PDF.

This can take tens of minutes on average for a human depending on the amount of data to check and the delay between each claim. Also, the process is limited to employee working hours.

In contrast, an RPA robot would perform steps (1) to (4) continuously without human intervention, unless it finds a discrepancy, in which case it would notify the human operator. Finally, the robot would send an automated email using a predefined template to the relevant department for further processing. This can only take a few minutes, is done exactly the same way every time and all day long, while providing an audit trail. Insurance companies can thus save on costs, and improve their efficiency while keeping human intervention for more complex business processes.

AI: Automation of Reasoning Tasks

Artificial intelligence “combines the human capacities for learning, perception, and interaction [...] at a level of complexity [and automation] that ultimately supersedes our own abilities.” It is a spectrum of technologies (e.g., natural language processing, computer vision, predictive modeling, data clustering, and many more) that opens new use cases for businesses, as well as reduces entry cost for many existing business problems that still require too much human intervention. 

A few examples of AI applications in business are fraud detection (predictive modeling and/or data clustering), facial recognition (computer vision), extracting topics from large text databases (natural language processing), recommendation systems (e.g., for marketing or e-commerce), etc. 

In order to tackle these use cases and leverage the benefits of AI in business, using a data science and machine learning platform is a best practice. — it is the key to successfully scaling AI projects and to bringing a robust data methodology to all levels of the business. Dataiku is a leading Enterprise AI and machine learning platform that not only allows organizations to tackle many AI use cases, but it also enables business people, analysts, and data scientists to collaborate around AI while also providing a sound framework for the governance and operationalization of these projects.

→ See what industry analysts Gartner and Forrester say about Dataiku

Example Use Case: Physical Mail Processing

Many companies still receive a lot of physical mail in various formats (typed, handwritten, forms, leaflets, etc.), and the processing is often outsourced to external service providers. This is both costly and not perfectly accurate, as some internal rerouting is usually needed. 

AI can help by automating one of the most time consuming processes: reading handwritten mail before routing it to the right department. This is done in the following order:

  1. Classification of typed vs. handwritten mail — typed mail is sent to OCR software (this is RPA!), while handwritten mail is further processed using AI.
  2. Detection of text areas, lines, then words.
  3. Recognition of handwritten words.

For (1), we must manually create a labeled dataset of typed, handwritten, and other mail, then train a machine learning model to classify between the three categories (for instance, using autoencoders to form a latent representation of the scanned mail). We then use various image decomposition techniques to perform (2): dilatation by convolution and connected components for text areas, and projection profile method for line and words. Finally, we perform (3) using deep learning (CNNs and LSTMs) in conjunction with a method called Connectionist Temporal Classification.

After automatically reading the mail, the resulting text can also be analyzed using Natural Language Processing (NLP) in order to automatically extract relevant information from it, and provide insights to improve the internal mail routing.

This use case was presented during the EGG UK 2018 conference:

 

When To Use Both

As you must probably understand by now, some business problems require both technologies to be successfully solved, the same way humans use both their brain and brawn in everyday tasks.

Here are two examples of combined AI/RPA usage:

  • Bank customer claim: First, the customer interacts with the bank’s chat bot to input its claim. This mainly leverages AI and natural language processing to translate the customer’s natural text query to structured data that can then be processed. In a similar manner to the insurance claim example above, RPA is then used to file the claim on the bank’s internal platform
  • Scanned PDF invoice processing: AI and image recognition algorithms are first run to find areas with relevant data in the PDF. RPA and optical character recognition are then used to extract the text from the relevant PDF areas. Then AI and natural language processing analyze the text and categorize it, for the RPA robot to finally input the invoice data in the company’s sales/purchasing software.

You May Also Like

Building a Culture of Experimentation

Read More

Building a Modern AI Platform Strategy

Read More

Why Is My Data Drifting?

Read More