Technologist, Futurist, Entrepreneur, Investor, Cofounder and CEO at intelligent automation company Rainbird Technologies.
Around the globe and across industries, inflation is driving up costs. More regulation means more compliance and more complexity in doing business. All businesses need to innovate to stay ahead but must also minimize the risk of innovation failure. Central to this is the desire to automate what can be automated but also deliver a highly differentiated service, which often means services with more (not less) humanity.
Because organizational decision-making is complex, however, many untapped opportunities lie within this process. Consider three different types of enterprise-level decision making.
1. Simple decisions. With this type of decision, the data is predictable, and the algorithm is known. These can be automated using simple workflow tools.
2. Complex decisions. There are more of these types of decisions than in the first category. Data analytics and other forms of data science can be used to present insights to human operators so they can make better business decisions.
3. Decisions that are complex, contextual and non-linear. This is the most prevalent category. Everyone has, quite frankly, given up trying to automate this type of decision making.
People have given up on this third category because the data is fluid, incomplete and often of poor quality. These decisions rely on humans with vast amounts of training and experience to make inferences in the absence of precise data. The added problem is that human experts often rely on non-quantitative factors, including personal experience, intuition and input from peers.
The third category represents a vast untapped opportunity to augment human expertise across more complex decisioning processes and radically change both the efficiency of operations and the outcomes for customers.
Knowledge graphs offer a new way to automate this type of decision making.
What are knowledge graphs?
Knowledge graphs complement the way we work as humans. We have knowledge in our heads which is the result of our formal education, our experience with others and the rules we abide by to do our jobs. We examine the data we have and ask questions when we need more data. Eventually, we have a good enough picture to draw inferences and make conclusions.
One way to think about knowledge graphs is as knowledge structures that use ontologies to integrate data. They are often used to describe concepts and the relationships between them to represent networks of concepts. As opposed to traditional linear-based modeling, knowledge graphs allow data to be organized into frameworks for integration and ensure understanding between the developer and the user.
They can be populated automatically using legacy systems or external data sources, but it is possible to manually encode human knowledge on these descriptive ontologies in a non-linear way. Inference algorithms can then be used to “reason” over these models and their data connections to derive high-quality answers to complex questions.
Despite knowledge graphs being well understood in academic circles, businesses have lacked the appropriate methodologies and tools needed to build and deploy reliable knowledge graphs in the enterprise. For too long, knowledge graphs have been underutilized, while workflow and decision-tree tools that are based on “if this, then that“ logic have become ubiquitous. Enterprise-ready tools that can leverage knowledge graphs have been lacking.
However, recent innovations in the intelligent automation space have seen new technologies begin to emerge. These new tools allow users to digitize knowledge to deliver efficiencies beyond the abilities of linear workflows and robotic process automation (RPA) tools. These tools have also delivered a level of precision and audibility not available through data science.
The capability is powerful because, as humans, we don’t think linearly but yet almost every tool we use, from the humble pen and paper used to write a list of instructions to modern workflow engines, is linear. But knowledge graphs can offer the opportunity to model our complex world in a non-linear way without needing to be a software engineer.
How do organizations get started with knowledge graphs?
So now we have the technology, what else is in the way?
History demonstrates that enterprises are slow to adopt new methods because “change” requires a strong will for things to be different, a degree of disruption and investment. Before investing in innovative technologies, organizations have to ask themselves: Is the juice worth the squeeze?
Organizations that are rich in intellectual property but constrained by the availability and cost of labor should be the first to consider this option as a way to alleviate overworked domain experts and expand business operations simultaneously.
But it is therefore critical that domain experts are central to the process of encoding their own knowledge and reducing over-reliance on engineers. Instead, engineers and experts should work in tandem to create usable and interactive computer models that prevent expertise from being lost in translation, unlocking new methods of efficient scaling.
Human beings are exceptionally good at creating ontologies, so building these systems is fast once the process begins. Computers are good at spotting patterns but deciding what is important is an intrinsically human task. Once you’ve done that, computers are much better at processing information and making complex decisions at scale.
The result is an ability to harness the complementary strengths of humans and machines to create a new kind of workforce that is smarter and more resilient than anything that has come before.