Nobel Laureate Daniel Kahneman, in his book "Thinking Fast and Slow” (2011), describes two ways of human thinking - System 1 and System 2. In short:
- System 1 is always turned on. It’s automatic, fast, unconscious and stereotypic. It is never in doubt, and extremely quick and efficient in recognizing known patterns and fitting the new concepts into the established perspectives. It is in use while you are driving on an empty road without thinking and can read text on billboards without paying attention. At work, System 1 makes you feel comfortable in familiar situations; however, it also leads you away from data-driven decision-making towards intuition and biases.
- System 2 is the opposite – it's usually turned off. It is slow, conscious, calculating and logical. You use it for complex calculations (e.g. 13x17=?), analyzing data and for learning new skills. The problem is that System 2 is lazy, and it takes focused efforts to turn it on. That is why people in many situations fail to use it and rely on intuition instead. Here is a famous puzzle from the book:
A bat and ball cost $1.10. The bat costs one dollar more than the ball. How much does the ball cost?
More than 50% of Harvard, MIT and Princeton students gave an intuitive answer - $0.10 and it’s wrong.
System 2 consumes a lot of mental energy, which affects human’s behavior and leads to the bad choices:
“While your attention is focused on the digits, you are offered a choice between two desserts: a sinful chocolate cake and a virtuous fruit salad. The evidence suggests that you would be more likely to select the tempting chocolate cake when your mind is loaded with digits. System 1 has more influence on behavior when System 2 is busy, and it has a sweet tooth.”
Today’s knowledge workers need to find the right balance of using System 2 for critical decision-making while not overloading it with unnecessary tedious operations. That’s why we happily rely on computer systems to operate numbers and other types of structured data. For people, dealing with numbers is not fun. It requires tremendous effort from human brains to turn on System 2 and check results to avoid errors.
It is especially unnatural, when the task requires you to accurately follow the standardized prescribed rules, while operating with numbers. Rules are simple, so the human brain quickly learns the pattern and then gets bored and switches to System 1. This helps with driving a car, but not with calculating numbers, which requires the full attention and active involvement of System 2. Constant switching between System 1 and System 2 is very tiring and leads to what’s called “human errors” - a deviation from the desired result, often due to stress, fatigue and reduced attention.
Here’s where computers come into play. They make calculations instantly with no errors and strictly follow the rules. That’s why modern organizations are looking to automate any process that does not require human decisions. Automation of standardized, rule-based processes that use structured data not only leads to cost savings, but also to better performance, higher quality results. It also leads to happier employees, who can now relieve System 2 from mundane operations and focus it more on “human” tasks – creative thinking and human interactions.
This is how RPA technology appeared at the edge of corporate automation. RPA bots are dream employees: they can log into enterprise applications using credentials, just like humans, and perform standard, repetitive tasks quickly with zero errors. Unfortunately, our world does not consist of purely structured data. In fact, for an average company about 80% of its data is “dark” data, not suitable for automation. This includes data from documents, emails, meeting recordings and other types of human communications. Processing this data requires human thinking to find the relevant information in unstructured formats and put it into a structure to feed standardized business processes.
This is where AI technology brings value. It can emulate “human thinking” and detect valuable pieces of data in semi-structured documents or completely unstructured text. Now, the “bat and ball” puzzle below can be solved by machine. It would apply computer vision, an OCR, to read sentences from the book, use NLP to parse them and identify objects/prices, and convert them into structured form:
|Bat + Ball||X+Y = 1.10|
|Bat - Ball||X-Y = 1|
Finally, it would apply standard algorithms to solve two linear equations and get the correct answers: Bat = $1.05 and Ball = $0.05, which means that the “human error” rate would drop from 50% to zero.
We envision the future of corporations where incredible productivity is powered by RPA bots that perform repetitive operations across all divisions of the enterprise and cognitive bots adjust unstructured, “human” information into a structured form. This approach will give human workers the time and opportunity to realize their creative potential working on projects and ideas that will improve our future.