Automation insights and productivity tips from LEAPWORK.

All Posts

AI and Automation: A User Guide

A lot of companies today have already adopted automation due to its ability to effectively perform tasks at high speed and accuracy. The next step is AI.

Adding AI to automation adds another dimension to the equation and lets these companies do so much more. It brings them towards hyperautomation which according to Gartner is “an unavoidable market state”.

But what is AI actually (in simple terms) and how can businesses use it in their automation efforts to achieve this “next-level-automation”?

In this guide we’ll give you a basic understanding of AI in automation so that you can utilize its capabilities and bring your automation efforts to the next level.

Watch our on-demand webinar on AI and Automation here.

New call-to-action

What is AI?

AI stands for Artificial Intelligence. The idea with AI is to replicate humans on a cognitive level and thereby to enable robots to think, speak and act like human beings.

If you were to compare AI to automation, you could say that where automation is about setting up robots to follow a set of pre-defined rules, AI is about setting up robots to make their own decisions (though still based on human input).

Today, however, AIs aren’t as smart as many make them up to be. What we have today is a junior version of an intelligence - also called Narrow Artificial Intelligence (NAI).

Siri is an example of an NAI. It can do things like tell you jokes or inform you about the weather. But NAIs don’t do much more than what they’ve been trained to do. They don’t have a conscience, they don’t have feelings, and they can’t think outside the box. In that sense they’re still very far from replicating humans on a cognitive level.

A better way to describe AIs today is perhaps that it’s a bit like a dog learning to give paw. You can teach it certain things, and at some point it will understand how to do those things on its own. What AIs do to learn is that they seek patterns, and based on those patterns, they ‘learn’ how to perform certain actions.

When explained in this way it of course sounds quite simple, but in reality it is a rather complex technology, and it’s extremely useful when used in combination with automation.

How is AI used in automation?

In many ways, AI perfectly complements automation: Where an automation tool can e.g. transfer data from A to B, the AI capability can interpret that data and respond to it (this is an RPA example, but AI is also very useful in test automation). For many businesses AI is thus a brilliant way to augment their automation robots’ abilities.

To understand how, it might be useful to look at some different types of AI.

Optical Character Recognition (OCR)

Optical Character Recognition (OCR) is perhaps one of the most utilized types of AI today.

OCR is a basic necessity for businesses that are going through a digital transformation because it allows them to move from paper to digital.

So what does it do? It essentially recognizes letters in a document, which could be either handwritten and scanned or computer-generated.

For example, let’s say that the HR department of a company wants to automate the process of entering data from signed contracts into a database. More specifically, they want certain details pulled from the document that they’ve scanned and entered into certain data fields of their system.

To do this, they first of all need computer vision, a component of OCR, to locate the text to be extracted. For example, it might locate ‘NAME’ and then pull the text that follows this text. Once the text has been extracted, it needs to be read. If the name is written by hand, the OCR technology will work to identify the characters, usually with a bit more difficulty. If it is computer-generated, it can read the text with higher accuracy. Once the text has been read, it can then transfer that text via an automation tool to the desired system.

how computer vision and OCR worksNatural Language Processing (NLP)

Natural Language Processing (NLP) is another frequently used type of AI. The purpose of NLP is to understand human language, as well as generate it.

NLP doesn’t truly understand humans, however. It understands a few basic properties of text. For example, in the sentence “What is the weather forecast for tomorrow?”, NLP will locate ‘weather forecast’ as the main entity of the sentence, and understand what the intent is (‘what is’) as well as place it in a context of time or direction (‘tomorrow’).

This is the technology that Alexa and Siri use when you ask them about the weather. In both of these cases, there is also another layer of technology which first translates speech into text and then text into speech in order to provide you with a response.

Companies use these technologies in several ways, such as for automated voice calls and chatbots. They can also use it in combination with automation to classify emails.

Voice calls

For voice calls, a speech to text technology is used in addition to NLP.

You might have experienced this technology yourself the last time you called your bank or mobile service company. First, they might ask you 1-2 questions to authenticate who you are, or to understand your intent. This allows them to give you a more personalized response and perhaps pass you on to a human agent that can solve more complex problems.


Chatbots serve a similar purpose, but don’t require any speech to text technology. They use NLP to understand a user’s intent, interpret it and send users down a certain path based on that interpretation.

Email classification

A last use case, which consists of both NLP and automation is email classification. This involves setting up a robot to monitor an inbox, and classifying emails based on certain criteria.

For example, a mobile services company might have thousands of customer support requests coming in every day, and therefore need an efficient way to sort and delegate these emails. Instead of just classifying the emails by a keyword (that could be done with automation), NLP can be used to interpret questions and respond according to a template.

Read how to do email classification with AI and no-code automation here.

All in all, these ways of using AI and automation can help businesses save time and resources and improve customer satisfaction levels significantly, as customers receive faster responses.

This is the reason why more and more businesses are looking into AI as a part of their digital transformation. But how are they doing this and what are the implications of AI and automation now and in the future?

How is AI in test automation and RPA different?

Above are examples of AI used in Robotic Process Automation (RPA). AI does, however also play a significant role in test automation and can add a lot of value to quality assurance.

To learn how AI is used in test automation, read this blog post.

If you’re unsure about the difference between RPA and test automation read about it here.

How can you get started with AI based automation?

As AI is becoming increasingly popular, many businesses wonder how they too can get started with the technology and implement it as a part of their digital transformation.

The challenge for these companies is in finding the right business cases for AI, rather than implementing AI just for the sake of AI (which, by the way, also is a valid point also when talking about RPA).

1.      Identify business need

Businesses must identify what it is that you cannot solve with your current set of tools and pinpoint the bottleneck. So instead of saying “everyone is using chatbots, should we too?” ask your customers for feedback on your services and investigate what the best technology is for solving that problem. Are they saying the response time is too slow? Then a chatbot using NLP might help you speed up responses. Or are they saying the quality of responses is low? Then robotic responses are probably not your best bet.

Only when you’ve uncovered the real customer needs should you start looking at the technologies.

2.      Research available technologies

Next should be a thorough research phase into which AI technologies to focus on. Here, you might want to bring in expert help as AI is an extremely complex field with many more subtechnologies than those we’ve mentioned above. Bringing in consultants will probably be worth the investment.

3.      Select your tool

Once you’ve decided on your technology comes another important research phase – finding the right tool vendor. No matter what your needs are, it will pay off in the long run to select a tool that will allow you to work across technologies, and that will enable you to integrate any AI capabilities, despite your existing tool landscape.

4.      Implement

Last is implementation. In this phase it can be a good idea to test your new technologies internally first; employees are usually more forgiving than customers. Start with the small problem areas and learn to solve those efficiently. Once you’ve aced this, you’re ready to scale it up.

Are you ready to use AI in your automation efforts?

Sign up for our webinar on AI and hyperautomation to learn more about the topic, to see a common AI use case, and to find out how you can get started with AI and automation fast and easy, without writing any code.

New call-to-action

Maria Homann
Maria Homann
Content Marketing Manager

Related Posts

Parallel Testing with Codeless Test Automation

The pressure on software development teams to deliver quality at high speed has never been greater. Teams must be able to test continuously and be agile in an increasingly digital world, driving the need for new methods of quality assurance.

Why Continuous Integration isn’t Continuous without Test Automation

Becoming agile will take time and require commitment, but it is nonetheless vital for harnessing the forces that are accelerating innovation and digital transformation for organizations worldwide.

A Guide to Achieving Continuous Integration in Agile

Becoming agile is never a straight path to success. Teams will prepare strategies, research and implement tools, adopt Scrum frameworks, and still struggle to achieve a truly agile software delivery cycle.