Software robots and people seem to be adversaries competing for work, or at least this is the perception when inexperienced RPA teams approach a company or department automation. When it comes to set-up the automation initiative, it is usual to separate processes into two groups: those executed by software robots and those people will continue to carry out. However, experience teaches us that the “all-or-nothing” vision may oversimplify business reality, preventing us from looking beyond the obvious to expand the range of automatable processes.
Therefore, it is likely a robot able to execute a whole process requires a human’s agent help or intervention, especially in business processes with dozens of criteria or exceptions. In such complex processes the key to success is to define properly what to automate to achieve a positive ROI in the shortest term.
In a RPA project, when a robot finds an exception or a new business rule not previously considered, two paths can be taken. The first option is someone “trains” the robot. In Jidoka, the trainers are RPA analysts and programmers working together to enhance the robot functional knowledge and to develop an improved version.
The second option is not implementing that exception or new business rule, and get an agent to solve it upon assistance request from the robot, which is known as a “cobot”: a collaborative robot.
Admittedly, today’s software robots do not make decisions autonomously nor they learn alone. There has been a lot of talk lately about artificial intelligence, machine learning, cognitive, but its implementation in real RPA projects is still small. The causes are diverse. One of them is that artificial intelligence projects require high investments (with a low ROI in the short term) and deployment that can last for months with a very low success rate. With these premises the automated process “business case” doesn’t seem very appealing, since it contradicts the benefits of a “pure” RPA project (aka “real”): low costs and rapid implementation.
Another reason is that AI, ML or Cognitive-based automation processes require highly qualified staff with proper programming skills and it is not always easy for companies to incorporate and retain these HR profiles (as we have mentioned in a previous post, RPA and HR talent might seem to be enemies…).
So, besides improving robots with new rules or exceptions, one of the most viable automation options in the short term is enabling robots to request help or assistance from human agents, getting them to collaborate with each other.
How do we solve this issue at Jidoka?
The Jidoka robots are able to identify not contemplated events or exceptions, and react to them requesting human intervention for their resolution.
The first method to approach exceptions is the visual recognition on the Jidoka console, as well as the configuration of an e-mail alert with the exception details, which will be sent to a business-profiled user so that he/she can solve the exception manually once the robot has finished the execution.
In addition, at Jidoka we have developed new methods and techniques to facilitate the robot-human collaboration, such as:
- A robot can request assistance, so a person can respond questions whose answers the robot does not know. This assistance can be provided using the Jidoka REST API so it can be easily integrated with a third-party application.
- A chatbot help the user through a private chat available on the Jidoka console, for answering the questions posed by the robot, interacting even in natural language (NLP).
These features can be implemented, for example, in the following use cases:
- A robot sends an alert to a mobile application, which, via REST API, get the question the person must solve. The robot pauses the execution waiting for the response of the agent. Once the exception or the question that only the human can answer is solved, the robot continues the execution.
- A robot warns the agent it needs help, for example, via email or SMS. The user accesses the Jidoka console where a chatbot shows the issue so the human can address the exception.
- A robot runs directly from the console’s chatbot. When necessary, the robot asks for input parameters for its correct operation and will return the results of the execution.
The practical application of such functionalities can be very broad, for example:
- The resolution of a CAPTCHA, where the agent must enter the data so the robot can continue its execution.
- Renting a car via chatbot, where a user asks, using natural language, for a car to book and the robot makes the appropriate questions to offer several rental options.
We have developed these use cases with Deloitte in Spain for their RPA Center of Excellence. This video shows such COBOTS in action:
Some may think this kind of human-robot collaboration is not innovative, as it is not the “robotic intelligence” so many people are talking about. And it is true, but it is practical and appropriate for the current RPA universe moment, when there are thousands of processes that can be easily automated with these functionalities.
In the years to come we will see the real impact of cognitive robots, that will be able to work with large volumes of unstructured data, and will help humans applying the power of business analytics in complex business processes. And also with an affordable cost reflected in positive “business cases”.
But not today, or is it that everybody already has R2D2 or C3PO in their homes and I haven’t noticed it?