What is RPA and how is it different from AI?

Let’s demystify Robotic Process Automation (RPA) and what and where RPA and AI are more applicable.

RPA is a software that copies human behavior or actions, whereas AI is a broad term where a machine simulates human intelligence such as sensing things, making deductions and communicate. The keywords here are – actions vs intelligence. However, many are mix up related jargons like  Intelligent Automation (IA) or Intelligence Process Automation (IPA). Fret not for this article should help you confidently differentiate between RPA and AI particularly in the context of automation.

IEEE defines RPAs as the use of preconfigured software instances that uses business rules and predefined activity choreography to complete the autonomous execution of a combination of processes, activities, transactions, and tasks in one or more unrelated software systems to deliver a result or service with human exception management.  Artificial Intelligence (AI) on the other hand is the combination of cognitive automation, Machine Learning (ML), reasoning, hypothesis generation and analysis, natural language processing and intentional algorithm mutation producing insights and analytics at or (hopefully) above human capability.

Simply put, RPAs are software bots that mimic human actions and AI is the simulation of human intelligence by machines.

Evolving from RPA to AI

RPA and AI are different ends of Intelligent Automation where AI is the evolution or the end of the spectrum handling increased complexities and cost. Let’s look at some simple examples to understand.

A classic example of RPA is invoice processing. Your suppliers send you electronic invoices by email, you download the invoices into a folder, extract the relevant information from the invoices, and finally create the bills in your accounting software. This is a great use case for RPA as it is suitable for automating these routine tasks.

RPA is suitable for automating the task of retrieving emails, downloading the attachments into a defined folder, creating the bills in the accounting software etc. These routine tasks can now be automated using bots.

On the other end of the spectrum is AI that can think and handle the tasks further. A great example of this is what Google does by sifting through your inbox. Google does a great job by retrieving the email based on the subject,  reading the relevant values such as invoice number, supplier name, invoice due date, product description, amounts due and more.

An RPA would not be able to handle this level of functionality as invoices are unstructured or semi-structured. Different suppliers have different invoice templates and formats. The item count and descriptions also vary. An RPA needs to be explicitly programmed or scripted thus making it impossible to teach the bot exactly where to extract the relevant information for each invoice. Certain RPAs are scripted to work alongside a human, acting as a virtual assistant easing the task, however expecting the human to resolve ambiguities. These are usually Robotic Desktop Automation (RDA). An RDA will pass the invoice through an Optical Character Recognition (OCR) to extract the required information and a human operator will validate this information before sending it back to the RPA bot to create the invoice in the system.

The upside of an AI solution is that the human intervention can be reduced to a minimum but the downside is the increased cost and project complexity compared to an RPA / RDA. From an implementation perspective, it is important to understand the balance of human to AI that the organization wants to strike. People-centric organizations that do not want to take the decision away from humans would rather have RDA.

Process-centric vs Data-centric

Another important difference between RPA and AI lies in their focus. RPAs are highly process driven, automating repetitive, rule-based processes that typically require interaction with multiple, disparate IT systems. At Accubits, for RPA implementations, process discovery workshops are usually a prerequisite to understand the AS-IS process and to document them in the Process Definition Document (PDD).

AI, is all about good quality data. This cannot be emphasized enough. The quality of the AI will depend on how we can train the machine learning algorithm with good quality and sufficient quantity of data.

Criteria for organizations suited for RPA implementation are:

1. Quick implementation and time-to-market (usually in weeks or months).
2. Low cost / budget.
3. The complexity of the problem is small, usually limited to a specific set of routine tasks without a deviation.

Criteria for organization suited for AI implementation are:

1. Massive amounts of data generated.
2. Unstructured or semi-structured data.
3. Highly complex problem but patterns should be discernible across different data sets so that predictions could be made or functions or operations could be automated.

Both AI and RPA are great digital transformation tools. The choice of implementing either RPA or AI or both depends on the specific use case and ensuring the strategic business fit.