RPA and Intelligent Automation – What’s the difference?

What is the “Intelligent” bit?

Simply marketing?

Scratching the surface of marketing can lead to definition wars. but examining some use cases for the technology will reveal the differences.

Gartner has the BOAT acronym for the vendor market, but separate “Hard Containers” for RPA and Intelligent Automation is probably not the optimal way to think of the technologies.

With different use cases for RPA and Intelligent Automation, is there a need to consider a unified architecture?

Any assumption that Intelligent Automation is Artificial Intelligence (AI) Automation is going to miss the point and leave generic marketing terms to hide the details.

Albert Einstein said:

“The measure of intelligence is the ability to change”

1. TLDR
2. Definitions for RPA and IA
3. Are Chatbots the added intelligence?
4. Are Documents the differentiator?
5. Is methodology part of the solution?
6. Could a “Best of Breed” approach be optimal?
7. Standards and the need for Orchestration
8. What is the future for Automation?

1. TL;DR

RPA is proven predictable automation.

Intelligent Automation encapsulates the use of AI into the processing.

Product evolution means that there are no longer clear boundaries in many scenarios.

Conversational interactions and document processing are where multiple technologies are often required.

Orchestration delivers the framework to enable “Best of Breed” architectures to be deployed.

The generation of automation code for deployment into an Orchestrated environment is at the early stages but is likely to be a key part of the future to reduce the development costs.

Improvements in the management of Orchestrated environments and more complex “Self-Healing” bots will drive down operating costs in the future.

2. Definitions for RPA and IA

There is no absolute definition for either term, but reasonable definitions here provide a basis for reference in this edition of the newsletter.

2.1 RPA

RPA which is an acronym for Robotic Process Automation. The term is used for software built to simulate the work a person would perform on computer systems in undertaking a business process.

Originally the distinctive feature of RPA was that it could interact with GUI screens just like a person would interact with the screens. This delivered consistency of actions in a non-intrusive way so that the existing applications did not require any changes.

RPA has evolved, with the technology able to utilise API interfaces. The equivalent of work a person would do through a GUI interface can be implemented with a series of API interactions performed by RPA software.

RPA is a “Rules” based solution which follows specific logical sequences of activity. This enables “Un-Attended” automation as the potential outcomes are known.

2.2 Intelligent Automation

Intelligent Automation sometimes abbreviated to IA, is the ability to use AI, Machine Learning (ML) and Natural Language Processing (NLP) to implement an automation of a process.

The AI can be the use of GenAI’s ability to understand text to extract information, its ability to use NLP for input instructions as well as a response, plus its ability to perform pattern analysis to deliver “Computer Vision”.

By implementing solutions where people can be part of a review process, “Human in the Loop – HITL”, it is possible to achieve ML for the AI model which is used in the processing.

IA is able to deliver a “Probabilistic” solution which can cope with un-structured or semi-structured text as input. Thresholds are used to determine when the outcome from the work needs to be reviewed by a person.

2.3 Gartner BOAT

Gartner’s “BOAT” definition stands for Business Orchestration and Automation Technologies.

 3. Are Chatbots the added intelligence?

Chatbots have been known for providing a text based conversation between a user and an automation for some time.

The ability of a Chatbot to understand text more broadly that simple keyword matching has been dramatically enhanced by GenAI.

The conversations with Chatbots can be much richer and create the impression that there is a level of intelligence determining the responses.

Although the language in the conversations has enhanced, chatbots are inevitably limited by the range of automations available to them as actions from the dialogues.

There are good business reasons for such limits on the automations just as there would be if a person was speaking with human customer service agent.

It is possible to think of a Chatbot providing the conversational interface to Intelligent Automation or RPA, as either could be the mechanism to implement the actions.

4. Are Documents the differentiator?

The handling of documents requires the ability to effectively extract data from an unstructured file and the ability to generate files with a variety of contents.

It is the GenAI’s ability to understand text which provides the extraction capabilities for common business documents such as invoices. The variety of invoice structure requires the intelligence of AI to recognise and extract the elements.

With invoices containing a variable number of invoices lines, potentially spread across multiple pages, it is easy to understand why a simple of set of extraction rules for RPA to follow rarely delivers a satisfactory solution. This use case is often used to illustrate Intelligent Automation.

To achieve effective automation of standard business processes typically requires Intelligent Automation delivered as AI and ML for HITL implementation.

Intelligent Document Processing (IDP) is a common requirement which has emerged on top of base technologies Optical Character Recognition (OCR), Image Mark Recognition (IMR) and image recognition (known as Computer Vision).

As the management of the processes handling documents is common, many RPA technology vendors have incorporated the HITL activity with OCR, IMR and Computer Vision into the products.

The leading RPA vendors have products which can utilise AI for document extraction, enable complete IDP solutions to be deployed.

With the exception of a NLP command interface, the leading RPA products can provide the complete scope of Intelligent Automation solutions.

Some RPA technology has been enhanced to use AI Computer Vision functionality to address a traditional weakness of implementations. RPA deployments that use GUI screens for interactions are often viewed as “Brittle” due to relatively small presentation changes in the GUI causing the robots to fail and need maintenance work to adjust them. This RPA enhancement is marketed as a “Self Healing” bot.

It is this breath of capability from the RPA technology vendors that has created the blurring of definitions.

5. Is methodology part of the solution?

With the range of technical capability, it could be the methodology or approach that is effectively the differentiator in solutions.

By using AI to perform “Understanding” at various stages in the automation of processes the concept of “Agents” can be created, and the category of “Agentic AI” is born.

Some AI gurus will say that in Agentic AI, the ability to “Plan” and “Reason” are used. It is one interpretation.

In my opinion, it is AI understanding text and using probability to generate a response just like it does in all conversations. Yes, there can be some effective use of memory and AI effectively doing recursive processing on its own output rather than simply using its first response to complete the conversation. However, it is the same handling of text.

6. Could a “Best of Breed” approach be optimal?

With both Intelligent Automation and RPA both incorporating a growing list of component technologies the questions about a Unified solution or a Best of Breed approach can be considered.

The GenAI world has evolved at pace with many different generic LLMs and an increasing number of specialist Language Models. The dynamics of such a market has created a demand for open interfaces to enable the potential use of different AI products for different solutions and with expectation that things will change over time.

As well as GenAI, there are a variety of OCR engines available, and many RPA products which can offer different integrations with a range of applications.

In large enterprises and in mid-market business there will be many applications deployed across the business, so it is likely that a range of tools will be deployed to implement different automations.

7. Standards and the need for Orchestration

To achieve effective integration between products that are utilised in an automation there needs to be standards for how they will interact.

Business Processes have a recognised standard for their modelling and description called Business Process Modelling Notation (BPMN). If all products in a Best of Breed architecture comply with BPMN, it is possible for effective interactions to be implemented.

With many transactions going through a business process which involves many different activities including some that require HITL work, the co-ordination and persistence of data, require significant orchestration to be used.

Orchestration can deliver the framework infrastructure to enable security, logging, audit, monitoring, etc. for the business processes automations to operate in a managed environment.

Orchestration is not just a passive set of standards, it can be used to control the dynamic allocation of resources to achieve a schedule, enable priorities to be applied and provide the resilience when something un-expected occurs.

8. What is the future for Automation?

Some elements within the automation arena are mature (e.g. OCR), whereas other aspects such as the AI generation of code for automations are relatively new.

Following Gartner’s definition of the BOAT market, means that vendors are likely to increase their product coverage to the complete range of functionality. The competitive pressures and growing capabilities of AI, are likely to deliver simplification of the effort to deploy the technology as well as more effective operations with reductions of tasks going through HITL activity.

Improvements in the management of Orchestrated environments and more complex “Self-Healing” bots will drive down operating costs in the future.

Although the BPMN standard exists, it is still likely that considerable effort will be required to switch between vendor’s products. The use of “Best of Breed” and heterogenous environments are likely to continue to be used in many organisations.

Whether the terms RPA and Intelligent Automation continue to be used, get combined into a term such as “Intelligent RPA” or whether they get consumed by a broader term such as “AI Automation” will largely be determined by marketing. The technical components will remain at the operational level whichever marketing label is used.