By Elad Tsur, CEO and Co-Founder of Planck
The digital age continues to transform our personal and professional lives, with new methods to connect, communicate, and commercialize introduced nearly every day. The data available through the world wide web and the internet of things (IoT) is growing exponentially, creating a constant flow of information that can be channeled to power data-driven business decisions. But with more data comes more noise, and manually plumbing the depths to find relevant facts in a sea of unrelated blog posts and images can be difficult — if not impossible. On top of that, in many situations there is no practical way to access all of the points needed to make a 100%-data-driven decision.
Business decisions can be informed by many factors, but should always be based on the reality of the situation. And that reality is revealed by examining data. But, when time and resources are limited and difficult choices come calling, there is a strong temptation — and sometimes a necessity — to complement hard facts with intuition.
It’s not entirely clear where intuition comes from. Intuition motivates decisions that usually cannot easily be explained to others; or sometimes even yourself. In ‘Gut Feelings: The Intelligence of the Unconscious,’ psychologist Gerd Gigerenzer promotes the idea of non-conscious processing, which suggests that “intuition occurs with very little awareness about the underlying cognitive processes.” The human brain uses shorthand equations to help process an extremely complicated world, creating an unconscious learning framework used to recognize patterns in complex systems.
So, where does human intuition really fit in the digital age? Startups rely on intuition for growth, as their market and business are often undefined as they work to establish a foothold. Fledgling companies have less history and experience, meaning less access to historical data. It’s possible to look to other companies for a comparative path forward, but each market entry is unique. Trial and error, success and extrapolation, failure and course-correction — continual data-gathering, analysis and application — are how intuition is trained and confidence strengthened.
When I started Planck within the commercial insurance space, there were many decisions to make for which there was really no precedent. I grew up in Israel, where my father was an insurance agent. It made sense to move into the family business, which was also the largest market available to deploy our innovative AI data platform. I chose a business partner, with whom I had worked previously at Salesforce, and we opened an office in Tel Aviv.
In hindsight, all these decisions make logical sense, and each has subsequently contributed to Planck’s current success. But at the time there were too many unknowns and not enough time to properly calculate, so I had to rely on intuition — from the very small insignificant decisions, to the cardinal ones. What are the commute times for an office in the city? How does the rent compare? I had experience with my business partner as an employee and a friend, but will his skillset translate as a company founder? Is insurance the most appropriate industry to develop this technology? These decisions were supported by as much data as possible but, in the end, we had to choose the paths in which we were most confident.
Another example was introducing Planck in Japan, which was also based on intuition. Our initial analysis indicated that the move would draw too many resources from our U.S. business, but the data was not conclusive to the point of taking the option off the table. I felt very confident that transitioning into the Japanese market could actually make the platform stronger by learning new insurance practices, thereby benefiting the U.S. market. And, eventually, it was found to be the correct decision.
To better contend with emerging business complexities, more and more business leaders are turning to artificial intelligence to process and understand large data sets to aid in decision-making. AI and machine learning models aggregate all available data to create new insights. Deep learning, a type of machine learning model, is a neural network containing many hidden layers comprising millions of single neurons, each receiving input from many thousands of other neurons. Much like human intuition, it can be very difficult to follow the input to the output and explain what happened because there are so many open variables and factors in the model.
Experienced people are often capable of making better decisions with limited input because their choices are informed by their history. Their sharpened intuitions allow recognition of patterns and pathways that may remain invisible to less seasoned individuals. For example, in ‘The Imminent Inflection Point in Motor Insurance,’ Dr. Coenraad Vrolijk draws from 20 years of automotive safety experience to predict a substantial decline in the price of car insurance.
AI collects and analyzes massive amounts of data well beyond human capacity, and can be effectively trained to replicate and apply ‘experience’ to create insights. Deep learning models use linear algebra to convert the input into applicable business information, and can even be used to predict future states. This extremely complex process may seem random to the human observer, but it is fully deterministic.
In terms of application, these ‘artificial intuition’ determinations must be applied judiciously — particularly within a highly regulated industry such as insurance. For instance, the Planck platform can determine that a restaurant provides deliveries, even though the company may not promote this service anywhere. Based on the menu, hours of operation, comparable businesses, and thousands of other data inputs, many times it is reasonable to make this conclusion and verify the risk data. However, a machine learning model used to generate an actual quote based on derived insights requires more explanation into the factors that influence the final numbers. Just like personal history can present bias and narrow thinking in human decision-making, it is vital to ensure the program is not using historical data to discriminate against any group of people or location.
Human intuition will always have a role in business decisions, but its overall effectiveness may be limited. Business leaders now have additional sources to temper intuition and guide more effective data-based thinking. A simple machine learning model can be quickly trained on millions of situations, limited only by the amount of data collected. If an AI is trained using the submissions of thousands of insurance underwriters, that model can now draw from those experiences. And the ‘intuition’ for this specific domain will most likely be more accurate than any individual underwriter.
Data collection, analysis, insight and intuition — human or digital — all work toward the same goal: building confidence to make the right choices. So, when validating your own intuition or the insights provided through a complex AI analysis, the question shouldn’t be ‘Why?’ or ‘How?’, but rather ‘How confident am I, and should I make a decision based on it?’
CEO & Co-Founder
Planck CEO and Co-Founder Elad Tsur is a serial entrepreneur who is passionate about bridging the gap between technology and business – specifically in insurance and other financial services. With Planck, he has helped create an AI-based data platform for commercial insurance that enables insurers to grow new and organic business while drastically reducing their loss and expense ratios.
Prior to founding Planck, Elad started BlueTail Corporation, which delivered the first comprehensive AI platform for customer relationship management. Acquired by Salesforce.com in 2012, today it is known as Salesforce Einstein.
In addition to his work with Planck, Elad is a mentor to Israeli entrepreneurs, an angel investor in AI and deep tech startups, and is a biometrics expert volunteer for The Biometric Application Commissioner (Prime Minister’s Office – Government of Israel). He received a bachelor’s degree and a master’s degree in computer science from Open University of Israel as well as a master’s degree in computer science from Tel Aviv University.