Now, if you haven’t heard of RPA yet, let us offer you a small intro. Simply put, RPA allows for the simulation of a human user’s actions. Think about that tedious task of switching from one application to another, copying and pasting data across numerous applications.
Configured correctly, RPA can also act on basic decisions e.g. IF this THEN that. RPA allows us to automate these processes in a quick and efficient manner, without the need for new programs and applications. The advantages and disadvantages of this approach are as such:
• Advantage: If the process is structured, digital and rule-based, RPA can do it (RPA has no issue accessing things like APIs as well).
• Disadvantage: The moment the process changes (a button changes position, an extra functionality is added, an update is released) the RPA solution needs to be updated and maintained, or it will stop working
Indeed, this technology can do amazing things, automating processes that shouldn’t need to be performed manually. It’s hard not to think “this can solve all my problems!”. Unfortunately, as is often the case with these types of things, it’s not the one-shot, cure-all we were all hoping for.
For instance, sometimes problems in RPA development can be caused by an underestimation of the contribution the user plays in the process, which often only becomes apparent after various sessions of analysis (e.g. complex decisions points that require human judgement). In short, even today after close to 20 years of existence, RPA is not nearly as smart as an actual person. Therefore, it’s important to manage our expectations with RPA when implemented solely on its own.
This is where Artificial Intelligence can bring so much more …
Where do we focus?
While A.I. is in fact an excellent path to augment RPA capabilities, the question quickly becomes: where in that vast sea of business process inputs do we want to focus this new-found intelligence? Thankfully we have the answer in the 3 points below:
• Structured and digital data (persistent data, SQL-accessible)
• Unstructured but digital data (texts, word, pdf, website content, photo, video, etc …)
• Unstructured and non-digital data (paper documents)
How do we handle each part?
Structured digital data
here, we can apply data science concepts and Artificial Intelligence to assist with decision-making and scoring: classifiers, anomaly detections, linear regressions, and more.
Doing so increases the scope of possible decision points that can be handled by an RPA solution (replacing decisions that were previously limited to human judgement), and allows for RPA solutions that cover larger parts of the process, or enable processes that were previously entirely out of scope.
Unstructured digital data
this data needs to be transformed into structured data. To do this, we generally use a combination of NLP (Natural Language Processing) and Image Recognition.
This solves an issue that plagues many potential RPA solutions: the input data is not presented in a structured, ready-to-process format.
Unstructured & non-digital data
A.I. can be applied to optimize the scanning process through configuration of scanners, OCR tools and document recognition.
This allows collection of an even larger set of data: paper documents, text written in cursive on documents, ect. ..
That’s all for now. Keep checking back, as Part 2 will delve into specific use cases.
Thanks for reading!
Authors: Jerome Fortias, Daniel Fastenau