Limitations of RPA
RPA, too, has drawbacks that have been pointed out by many. Some of the most common limitations of RPA are:
RPA cannot read incomplete, non-electronic, and unstructured dataimagine receiving a paper-based document or a handwritten invoice.
In such case, RPA would not come in handy, since this technology can only automate processes in digital formats. It cannot scan handwritten notes (let alone decipher them) and it cannot automatically acknowledge if vital information is missing.. In this situation, using people or integrating a digital capture software would be more efficient (yet, costly).
RPA cannot execute decisions based on intuition and acknowledge exceptional scenarios
Imagine that your supplier sends you an invoice with a computational mistake but calls/emails you right after to inform you about it instead of re-sending the corrected file. In such case, the bot wouldn’t be able to acknowledge the informal agreement between the two parties and would execute its task incorrectly. Therefore, if we rely too heavily on RPA, we might unintentionally miss such human communication, which would inevitably lead to greater problems.
Making changes to the process you are automating will require reconfiguring the RPA bot
This applies more heavily to programmable bots rather than intelligent ones. For example, if you have programmed your bot to:
immediately analyze the invoice
open the billing software
input the required information
and store it in the correct folder
The same will not apply if you want it work identically with other types of documents too (for example, orders from clients). In order for the bot to carry out a wider variety of tasks besides the one it was programmed for, you should reconfigure its functions each time you need it to do an extra task. To avoid this, you might want to program a second bot, which, in turn, would be an extra cost.
Even though RPA is an extremely cost-sufficient technology, there will always be work processes which require human touch. Thus, integrating bots and smart machines might not always be the most cost-efficient solution.