You don’t need me to tell you that two of the biggest technology trends in business right now are Artificial Intelligence and Robotic Process Automation (AI and RPA). But if I said to you that they were Intelligent Automation and Cognitive Computing, or Service Delivery Automation and Autonomics, and that these were actually pretty much the same thing, then you might start to question me. I could even take it further and combine both of these into a super-trend of Cognitive Robotic Process Automation, at which point you would have given up all hope in me.
But this is exactly what is happening in the world of technology marketing at the moment. Just when you thought the industry had finally settled on some common terms for something, somebody muddies the waters by inventing a new one. And not because it describes something fundamentally different, but just so that their offering or product sounds different from everyone else’s.
Now, this obviously isn’t a new practice, but the danger now is that the technologies we are talking about really are game-changers, and the need for clarity is crucial. Buyers need to understand whether they are getting, for example, a genuine Robotic Process Automation capability or a macro-driven piece of software that has been rebadged as RPA. They need to understand whether the Artificial Intelligence software really does self-learn or whether it is some convoluted logic dressed up as AI. Knowing these things will make the difference between success and failure of the project, and whether that investment that you worked so hard to secure will actually deliver the benefits that were promised.
So, here’s some pointers that will help you cut through the marketing hype and identify ‘real’ RPA and ‘real’ AI applications.
Robotic Process Automation
The term “robot” is useful here because the software replaces (or enhances) the work that a human being would normally do. Process automation has been around for a long while (even something like SAP can be described as process automation software), but the difference with RPA is the focus on the human tasks. RPA software’s real value is because it works at the “presentation layer” (the user interface) of the vast majority of different types of computer systems and can be trained to access and write to them relatively simply. This sort of “simple complexity” hasn’t been available before.
It is important to remember that the RPA software robots are effectively dumb: they will do exactly what you have trained them to do, 100 per cent of the time. But there is no “intelligence” in them. So, if anyone talks about Intelligent Automation, or Cognitive Robotic Process Automation, then start putting on your cynic’s hat.
The opportunity for obfuscation with AI is enormous, and many people have openly taken that opportunity. The challenge comes because there is no single definition of AI – my favourite is that it is any technology that is 20 years from fruition. But if you think of AI capabilities in three different categories, then it should become somewhat clearer.
Firstly, there are AI technologies that are great at capturing information. This could be done through Vision Recognition (e.g. recognising a face in a photo), Sound Recognition (e.g. transcribing words that someone is saying), Search (e.g. extracting data from unstructured or semi-structured documents) or Data Analysis (e.g. identifying clusters of behaviours in customer data). The first three of these require what is called Supervised Learning, i.e. they require large data sets to learn the necessary patterns, whereas the fourth uses Unsupervised Learning, which means that it can come up with the answers without you telling it what the question is. But all of these essentially turn (unstructured) data into information, and this is the most mature application of AI in business today.
The second AI capability turns that information into something useful: it works out what is happening. This is done through Natural Language Processing (e.g. extracting the meaning from an email), Reasoning (e.g. how should I act based on the information given) or Prediction (e.g. predicting buying behaviours based on previous purchases). Some of these applications, such as Prediction, are more mature than others, but all of these can provide real value to a business.
Finally there is the capability to understand why something is happening. This area of AI feeds off most of the others I have mentioned. This is the least advanced area of AI and is not yet relevant to business applications, but will obviously have a huge impact once it does.
All AI applications will fit into one or more of the above categories. If what is being described to you feels like a bit of a round peg compared to my square holes, then you should start to question those capabilities. And if the talk is all about “neural networks” or “machine learning” (both of which are underlying AI technologies) then simply seek to understand what it does, rather than what it is.
RPA and AI are two very different technologies and should be treated as such (if you remember one thing make it this: there is no such thing as Cognitive Robotic Process Automation). Each of these technologies do complement each other very well (for example with AI provided structured outputs from unstructured inputs, which can then be processed by RPA) which is why they can be deployed very effectively together. But, please don’t get fooled by the hyperbole of marketing speak that surrounds these – seek to understand or, if it’s still not clear enough for you, seek advice.
This article first appeared in sourcingfocus.com.