RPA is a current hype and ghosts through the companies. The first "low hanging fruits" are quickly harvested. But what comes next? In this article, we would like to show how the journey can continue for companies and illustrate practically how AI functionalities can be easily integrated into an RPA system.
With RPA, initial successes can be achieved quickly today
Many companies are starting an RPA project today. The fast ROI and the ease of implementation are the key drivers. RPA directly empowers business departments and the IT department to think about automating simple to complex use cases. It doesn't matter which business processes or which applications are to be automated. Use cases can be found in almost every area.
However, the degree of complexity must be taken into account. This is usually measured on three levels (light, medium, heavy) and is determined by, for example:
- Number and quality of IT applications
- Authentication types e.g. two-factor
- Number of activities
- Number of repetitions and loops
- Use of many different triggers
The complexity has an impact on the development time of the bots and therefore, of course, on the business case. There are bots that interact with the user, or those that work everything away in the background.
For some customers, a higher level of complexity is the use of artificial intelligence (see Figure 2). Here, a distinction can be made between reliable, existing AI models and the development of own AI models.
Companies are able to automate cognitive tasks and human decisions with the help of AI methods. Many people talk about the future here, but we are already automating processes with intelligent bots today and would therefore like to show the general functionality and the advantages when RPA is combined with AI as technologies.
Companies place themselves in an expensive dependency with the commercial AI platforms
Artificial Intelligence (AI) can be integrated into almost every IT product and service today. Today's application fields for the use of AI are e.g. facial recognition, recommendation engines, fraud detection, chatbots, voice controls like Siri, Alexa or automatic translations. Both "supervised learning" and "unsupervised learning" are used.
Companies are challenged to find use cases for AI and develop solutions for business. It can be observed that companies are starting larger projects. Many corporate leaders think big, want big, and are engaged in Big Data and related data analytics. They try to find out unknown data correlations in order to be able to use them for "intelligent decisions". Thus, the machine would be even more intelligent than today's employee who does not have this knowledge.
It is not uncommon for projects to be expensive, take a long time, or fail before the end. In this context, commercial AI frameworks are also a major cost driver. Companies try to achieve a breakthrough with an AI platform from e.g. IBM, Microsoft and thus place themselves in an expensive dependency.
Fast and successful: With RPA plus AI, intelligent automation can be used productively
With the interaction of RPA and AI in general and this article in particular, we would like to show that AI-based solutions can also be implemented quickly and cost-effectively for productive process automation. Bringing the two technologies together is already envisaged in existing RPA platform solutions. For example, UiPath or Blueprism have functions for integrating cognitive activities such as text analysis, sentiment analysis, and translation. These are based on the APIs of Google, Stanford, IBM and Microsoft and thus use these services, which are associated with appropriate licensing.
However, there is also the possibility to integrate own scripts via programming modules. The Python programming language stands out with its variety of AI libraries, most of which are available for free use.
Here, we want to focus on the above-mentioned cognitive functions and implement them in the RPA system with open source products.
Cognitive automation example with UI Path and Python
The integration of AI is easily possible according to the methods presented above. Python's wide range of cognitive features enables rapid implementation. With the multitude of Python AI libraries, various application scenarios such as machine learning search, data pipelines for machine learning, data mining, neural networks, semantic analysis or algorithm verification can be realized.
In the following example, the library "TextBlob" is used. It uses the Natural Language Processing Toolkit, which, among other things, can be used to estimate the mood.
In the showcase, a user can enter a text, which is translated into English if necessary. By integrating a Python script, the mood is analyzed and output.
The following showcase is intended to demonstrate how a very simple implementation of Python AI functionalities can be implemented in RPA scripts using the direct integration of these scripts.
One application scenario is the analysis of social media channels (social listening) and the monitoring of new product or service launches. Market sentiments are recognized directly, and the company can respond to them immediately. Likewise, sentiment analysis is used in customer service to sort customer inquiries in free text format and also to achieve a fast response time to customer feedback.
Currently, cognitive bots also allow standardized processing of requests by analyzing human free text and identifying the primary concern. Thus, unstructured information becomes data that is processed in a structured manner.
The use of AI can be implemented easily and quickly with the help of RPA. In addition, the combination of technologies opens up the respective area of application, i.e. there are more use cases and thus a potentially higher degree of automation. The advantage is that these solutions can be developed very leanly and in a short time. Existing AI models are used directly and new AI models can be developed with a limited database that fits the process. Thus, AI can be deployed productively in a short time. Potentials can be developed in an agile manner without having to set up large projects. Even companies that use RPA as an interim solution benefit from the use cases, because once deployed and well-functioning models can also be integrated into a long-term automation solution without any problems.