A week after the publication of an article in the New Yorker by Atul Gawande on the importance of coaches for surgeons and other professionals in fields that don't normally use coaches, the New York Times ran an article that raised the question whether elite athletes really need coaches at all?  

The Times used as examples a few extraordinary self-trained athletes to make the point that it is possible to compete at the top level of a sport without a coach, but never really explored whether any of their examples could still significantly improve their performance by working with a coach. I'm sure there are coaches who could offer insightful opinions on these individuals and perhaps constructive criticism.

In his article, Gawande noted that as an accomplished tennis player who was well-coached throughout his youth as a competitive player, he was surprised when a single session with a new pro after going coach-less for years led to a simple change that improved his serve speed dramatically. As Gawande points out, however, coaches can also actually hold back athletes, particularly those who take an unconventional approach to an activity such as Dick Fosbury, whose Fosbury Flop revolutionized the high jump against the advice of his coach.

How, then, does a coach enable a performance improvement and can this role be automated? Good coaches know what has worked in the past and can relate those experiences to the person being coached. I think of their personal store of knowledge as a collection of success patterns. They can spot deviations from these patterns in their ongoing evaluation of an individual or team performance. The pro who improved Gawande’s serve only saw him once, but that was enough to detect a misfit with his pattern for a good serve and affect change.

Good coaches also know some things that haven’t worked and can hopefully prevent common mistakes from inhibiting performance. What they cannot do, generally, is identify a new pattern based on the past and contextual knowledge. Fosbury's coach shouldn’t be faulted for not recognizing the Flop as a revolutionary advance, although perhaps more experimentation should be encouraged by coaches in general.

In athletic pursuits where the smallest behavioral changes can make the difference between perfection and bronze, I think it is probably unreasonable to expect an athlete to identify and correct their own issues. I remember watching Mary Lou Retton score a perfect 10 on the floor exercise and vault in the 1984 Summer Olympics in Los Angeles. Her coach, Bela Karoli, could no more have performed that routine than I could, but he brought out the best in Ms Retton. Top performers may know what they should do, but not be aware of what they are actually doing. That’s where coaches come in to play and as Gawande observes, that type of feedback shouldn't be reserved for athletes.

In the future, however, I expect coaches for athletes, surgeons, generals in battle, and many other endeavors will either use smarter computing techniques or be replaced by computers altogether. A computer with access to big data and analytics can perform many of the critical tasks coaches perform better than humans. They can store the entire recorded history of previous performances classified by patterns, or in an unstructured form and derive patterns on the fly.

In the mid-1980s I worked on a preliminary design for the Air Land Battle Management system. At the time, it was a state of the art project designed to capture historical battle knowledge and make it available to commanders in the field as a decision support aid using a set of cooperating expert systems. Today, such knowledge capture and management is only the beginning. Advanced analytics may enable knowledge extension or the identification of previously unknown patterns to guide action. The more things change, the more we need coaching for critical roles to leverage more data. When coupled with simulation and modeling capabilities that are fast enough to evaluate alternative patterns, the computer as coach will enable performance at levels previously unattainable for such tasks.

If the outcome of a complex human activity has the potential for significantly adverse consequences to a third party, as in the case of a patient undergoing surgery, or an army protecting or taking territory, I want to know that they are working with the best data available when decisions or adjustments are made. Frankly, I am not concerned about who wins the Boston Marathon, nor too concerned about who wins the World Series (once bad coaching and/or performance eliminate the Boston Red Sox) nor whether every performance by a concert violinist is their best possible work. Variations make life interesting.

I am, however, concerned when someone gets on a plane, goes under the knife, or into battle for me, that their pilot, air traffic controllers, surgeons, and officers are operating with the best possible information. That information should include historical data and timely feedback about misfits with desired behaviors to achieve the best outcomes. For these roles and others, the question becomes: Will a computer like Watson be the coach, or the coach’s coach? It’s only a matter of time. And patterns.