You Can Design Your Business To Prosper, Rather Than Suffer, From The Robot Revolution

Despite what we hear from certain prominent politicians, most serious commentators tell us that robots, much more than imports, are the main root cause of disappearing jobs. [I use the term robot broadly to include software based on artificial intelligence (AI) and machine learning.] If your job can be performed by every-more-capable robots, you are at risk. And, if your business is based on jobs that robots can take, you are at risk too. Your margins will be squeezed and you will be forced to push out a lot of employees, or lose to robot-enabled competitors. It’s time to evolve your business model to one that makes robots a force multiplier.

An analysis of the impact of AI on medicine provides rich examples of how this can work. Siddhartha Mukherjee, who is an M.D., a professor at Columbia Medical School and an accomplished author, describes in detail how artificial intelligence is impacting two medical specialties: radiology and dermatology. This essay, A.I. vs. M.D., was published in both the Annals of Medicine and The New Yorker. Radiology and Dermatology are high-powered fields: they are #6 and #8 respectively in ranking of MD incomes with average salaries above $300,000 per year.

Dr. Mukherjee notes that the core diagnostic task for both specialties is to distinguish diseased tissue from normal tissue by examining visual information: diagnostic images for radiologists and skin for dermatologists. In simple terms this means determining what the doctor is seeing, and what it means for the health of a patient. These tasks, recognizing patterns in visual information (“what I am seeing”) and, having classified the object, determining its significance based on the accumulated body of medical knowledge (“what it means”) are exactly what AI does better and better. AI is already out-performing human doctors in certain diagnostic situations, and it is not hard to envision that in the near future it will be equally or more capable at delivering diagnoses in a large percentage of situations.

The key question is: what does this mean for the individual doctors and the businesses that employ them? Many think the future for radiologists will be difficult. Geoffrey Hinton, a computer scientist quoted in Dr. Mukherjee’s article, puts it rather bluntly: “I think that if you work as a radiologist you are like Wile E. Coyote in the [Road Runner] cartoon … You’re already over the edge of the cliff, but you haven’t yet looked down … I said this at a hospital. It did not go down too well.” Hospitals that employ radiologists may see this differently, however. If scans can be obtained faster at lower cost, it will help other doctors give better care and eliminate unnecessary procedures. By comparison, Lindsey Bordone, a dermatologist quoted in the article, welcomes AI with a shrug: “If it helps me make decisions with greater accuracy, I’d welcome it.”

Dr. Mukherjee points to two key differences between dermatology and radiology. Radiologists are mainly concerned with “what” they are seeing and its medical significance. They typically have little involvement with why something happened or what to do next. Dermatologists provide diagnoses, and they are also concerned with the cause of skin disease they are seeing, because that often indicates a path to a cure. AI systems, while very good at making classifications, provide no information that indicates how they come to a conclusion, and hence they shed little light on the cause of the condition they identify. They are deep black boxes.

And, dermatology is a very personal type of medicine. Some problems are better identified by touch than sight, so a skin exam typically includes a laying-on of hands. Dermatology affects one’s appearance: a very personal matter. The discussion with the doctor is often about what caused the problem, and what behavior change (as well as medicine or procedure) can cure it. Many blemishes are benign and many common problems can be helped, so the experience is often reassuring. And there is often a follow-up: e.g., a person with a skin cancer diagnosis will be followed for life. It’s not surprising that dermatologists see AI as a tool to help them practice better, not a threat.

These examples offers lessons for every business. AI is becoming good at many cognitive tasks such as lesion classification that are high on the human skill hierarchy today. But AI based on machine learning, the main force driving the current AI revolution, is not good at determining cause and effect. AI has made inroads in aspects of human interaction such as cognitive behavioral therapy. However, I’m confident that humans will outperform AI in most complex, sophisticated human interactions, for a long time and maybe forever.

So if your business is based on developing rich relationships with customers, and helping them manage key parts of their lives and businesses over time, then you will probably find AI to be a force multiplier, not a threat. This is the dermatology example. If your business is knowledge-based (even deeply so) but transactional and impersonal, AI may well take over: e.g., radiology. Similar change has already happened in securities trading and routine customer service. And if your business is a mix of the two, you need to be shifting your value-add to the relationship-based activities, and taking full advantage of the AI revolution in the transactional activities. This is how hospitals should see radiology.

Viewed this way, the AI revolution is déjà vu: another wave of commoditization, like many in the past. Companies that survive commoditization and prosper do so by building strong customer relationships getting ahead of the change in their cost structures. Companies that don’t get squeezed. This is nothing that cannot be overcome. Time to get to work.

First posted @ on April 11, 2017.

Show your support

Clapping shows how much you appreciated Todd Hixon’s story.