Framing the value for cognitive computing
We have seen that cognitive computing represents perhaps the most complex toolset that humankind has yet to invent. We have seen glimmers of great gains that could be made in many professional fields and across a spectrum of consumer activities if we could build cognitive applications to support them. We have seen decision makers begin to weigh the impacts that such cognitive applications might have on their operations: in medicine, in marketing, in transportation, in insurance, in investing, in consumer finance, in entertainment, in the military—the list goes on and on.
We are anticipating a world that will operate very differently a decade from now. But however heavy the weight of the changes cognitive systems will bring to our overall experience, today’s decision makers need an analytic framework through which to evaluate the value proposition for any specific cognitive undertaking.
The Cognitive Computing Consortium has been doing research to provide support for an analytic framework for cognitive application decision-making. In concert with research partner Babson College and its information and technology management program, we are currently articulating a framework designed to offer executives who are facing opportunities or challenges in the world of cognitive computing a systematic way to consider the key decision elements for a particular application goal.
Consider an apparently uncomplicated example. How much might adding a “cognitive sales assistant,” a digital sidekick for our direct salesforce, be worth to my business? Start with a top-performing, “coin-operated” salesperson who has never heard of a cognitive sales assistant. What is the ROI associated with her? Count up the revenue she generates, subtract the exorbitant cost of her salary and commission, and you can run an ROI calculation on the value of her efforts. (At this point, let’s not complicate the analysis with how much additional revenue will be generated by not-quite-so-talented salespeople improving their performance by learning from her techniques and her example.)
Now compare that first analysis to a new analysis based on what might happen if you add a cognitive sales assistant to collaborate with this top salesperson on the job. Do a pilot study for a few months on the top salesperson’s performance using the cognitive assistant machine. Then count the beans again, and if the revenue goes up and the salesperson has some good stories about how the cognitive assistant helped her gain some new insights and close new deals, you should be able to run a second ROI on the financial impact of adding the cognitive sales assistant to the mix. In those two analyses, you can work almost exclusively with “hard dollar” financial streams. You will have no problem filling in the cells of the spreadsheet model you have no doubt created around revenue levels and costs. Simply introduce the cost of capital and present a decision maker with a tightly reasoned analysis.
But what if the lead executive should ask: “What will it cost us to make a cognitive sales assistant that is good enough to supercharge the results of even our top performers?” Suddenly you might have to consider what value to attach to the process of converting the wisdom of a top salesperson into a computer program, even if that program is as intelligent and high performing in its way as top sales professionals are in theirs.
Those of you who have been around knowledge management for a while will of course recognize this question as one that has dogged the field since the beginning of KM. Stated in its more traditional, general form, this is the problem of assigning a value to the process of transforming tacit knowledge into explicit knowledge.
Vehicle for disruption
Fortunately, we don’t necessarily have to answer that question directly. With the emergence of cognitive computing and the deep transformations that industry leaders are anticipating for their fields, many decision makers understand that their intuitive insights into potential impacts on their industries are strong enough to augment the traditional financial decision models. When your world might be disrupted in the near future, it’s worth exploring what can be done to evolve with it.
For example, decision makers can look at the newspaper business and understand how the Internet destroyed the value propositions of both the classified advertising business and the flow of recurring subscription revenue in that business model. They can look at taxi companies and other kinds of transportation businesses and see how the arrival of GPS, the new availability of consumer data, a transformed labor model and a group of aggressive new companies has crippled the industry’s traditional value measures. We could cite many other examples. The point is that cognitive computing has the potential to provide the vehicle for similar disruptions, and executives know that they will have to develop a textured understanding of what cognitive technologies might do to their field.
An emerging pattern
So the biggest challenge in understanding the value of a cognitive project is to delineate or establish a horizon for what the “soft dollar” investment might possibly be worth. An executive could ask: “What’s the value of the innovation group we set up?” Or she could ask: “What’s the value of improving the quality of our interactions with customers through introducing cognitive assistants for our customer service people?” Or, focusing on a big picture view of cognitive capabilities, she could ask: “What’s the value of anticipating industry disruption and realigning the business in advance of the shakeup?”
As a structured approach to those decision problems around cognitive systems, we see a pattern emerging in our research. Executives will depend on their analysts to cycle through a repeatable process: First, identify very specifically the particular value proposition—in soft and hard dollars—that the system will deliver. Then characterize the behaviors the system will exhibit to its users—the interactions and outputs available for human professionals. Then define the data landscape and concrete data elements that will be required to support those behaviors. Finally, select the technology components that will most efficiently operate on the required data and deliver the desired behaviors in a comfortable human interface.
This framework is not so far removed from the way systems development professionals are used to thinking about their projects. But in this case, getting the subtle adjustments right offers serious leverage toward the success of the cognitive project. And that project could be a key step in the process of preparing the business for a disrupted and unfamiliar future.