AI’s Olympic Gold-worthy Goal-Driven Systems
One of the most important things we can learn from Olympic athletes is the concept of focus, and how to set goals. This is more than just the idea of a long-term, ethereal objective that can take years to achieve; it is about making concrete goals, such as when a 100-meter sprinter crouches into the starting blocks before a race and closes his or her eyes to block out all but the most important stimuli at that moment — the gunshot that signifies the start of the race.
Many sprinters close their eyes because they fear being distracted by a competitor’s flinch. They know jumping out of the blocks too soon means a false start and instant disqualification. So, they shut out all the senses but the one needed most. Once the starting gun fires, the sprinter races out of the blocks, with eyes popping open, and then he or she focuses like a laser on the finish line, their new goal while blocking everything else out.
In her article Eyes on the Prize: Understanding the Links Between Perception and Motivation, behavioral scientist Emily Balcetis describes a series of experiments she dubbed ‘eyes on the prize’, which tested a strategy that motivated people to do something that might seem difficult and insurmountable. This amounted to people trying to improve their exercise routine and they were told to look at the distance to a finish line and focus their gaze solely on that line while ignoring everything else. The baseline group was told to act naturally. When both groups raced, Balcetis discovered the “people who kept their ‘eyes on the prize’ said the exercise required 17 percent less exertion than the baseline group. It hurt less. And they walked 23 percent faster.” Balcetis concluded that simply changing how people look around while walking improved the quality of their exercise and made the goal seem more attainable.
Both of these goal-oriented concepts reminded me of how Artificial Intelligence (AI) works. AI is a problem solver above all else — give it a goal and let it try to work out a solution. As Ron Schmelzer explains in his Forbes article Amazon Dives Deep into Reinforcement Learning, “Rather than being given good examples or discovering patterns on its own, reinforcement learning (RL) systems are given a final goal and learn through trial and error, discovering the optimal solution or best path to a goal.” Unsupervised learning can be used to find hidden patterns in data, which can be considered a goal-in-themselves.
When compared against normal rules-based analytics, AI truly shines. In its article Artificial intelligence Unlocks the True Power of Analytics, Adobe reveals the vast superiority of an AI-powered way over rules-based analytics. For IT departments trying to keep extremely complicated networks and systems operational, root cause analysis is extremely important. A rules-based analytics system would manually investigate why a negative event occurred and then consider all possible actions, while an AI-powered analytics tool, like AI for Operations (AIOps), would automatically evaluate what factors contributed to the event and suggest both the cause and an action directed at fixing the problem. The goal of an AIOps solution is to keep an IT system operational, often acting in a self-healing way so that the system is fully optimized and constantly learns about itself to keep it functioning properly.
AI is also useful in marketing, particularly in regards to campaign effectiveness, contends Adobe. The old rules-based analytics method means a business manually sets rules and weights to attribute the value of each touch of a customer journey that might lead to a conversion. With AI-powered analytics, however, an AI analytics tool automatically understands allthe factors that lead up to a successful marketing outcome and then attributes credit to each campaign element accordingly.
In marketing, the question for AI to solve might be, “How can I make a marketing offer so tempting it will not only be opened but also used?” Now, the variables here would be, “What is the best offer for the customer?”, “What is the most opportune time to send the offer?”, and “What is the best channel to send it on?” AI can help create a detailed understanding of a client’s purchasing habits to ensure the best possible offer is made. Once that’s done, an AI-backed marketing system, like Adobe’s Experience Cloud, Salesforce’s Marketing Cloud, or Pega Platform, needs to ensure the best possible offer is being received by the customer at the time when it’s most likely to be used. There’s no point sending out an email in the middle of the night if it will be lost in a sea of emails that need to be replied to in next morning.
An AI system can track when a customer normally opens emails and correlate those email opening rates and sales through rates. Social media can even be included into these marketing calculations. If a customer has a penchant for opening emails when he or she tweets, the AI system can be set up to look for customer tweets. A marketing email would be triggered once a tweet goes out. Of course, the sale is still far from being made, but the system’s AI is ensuring the sale has the best chance of occurring. Oscar Wilde’s most famous quote is probably, “I can resist everything except temptation,” and AI is basically understanding and even fostering that temptation to help make the sale as likely as possible, potentially with the best terms possible as well if a revenue optimization element is added.
In terms of customer churn, Adobe argues that rules-based analytics will try to manually spot patterns from reports on groups of customers who have defected. AI-powered analytics, however, will automatically identify which segments are at the greatest risk of defection and help organizations proactively stop this customer churn possibly long before a customer will be contemplating it.
When it comes to an organization finding its best customers, something the Pareto principle’s 80/20 rule shows is paramount for customer success, a rules-based analytics system will manually analyze segments in order to understand what makes high-quality customers different. However, an AI-powered analytics solution will automatically identify statistically significant attributes that high-performing customers share claims Adobe. Detailed customer segments can be created from this knowledge, which will give the organization data on how to handle each client individually.
In her Forbes article The Seven Patterns of AI, Kathleen Walch states there are seven commonalities to all AI applications regardless of the application. These are “hyperpersonalization, autonomous systems, predictive analytics and decision support, conversational/human interactions, patterns and anomalies, recognition systems, and goal-driven systems.” For Walch, the ‘Goal-Driven Systems Pattern’ is the seventh and final AI pattern.
For decades, Walch explains, machines have been beating humans at easily-conquered games like chess and checkers. Now, with AI-reinforced learning models crunching numbers on much more powerful hardware and sophisticated software, machines can take on and beat humans at some of the most complex games around, including Go and multiplayer online battle arena video games like Dota 2. “AlphaGo and AlphaZero were created by Google’s DeepMind division under the theory that through goals, computers could learn anything through game play,” explains Walch. Today goal-driven system patterns aren’t just for games, they can help with resource optimization, supply chain management, warehouse and logistics operations, iterative problem solving, as well as bidding and real-time auctions. Although the goal-driven systems pattern isn’t as widely utilized as many of the other AI patterns, it is gaining traction.
Eyes on the Prize
As in every Olympic race, time is of the essence. For those unwilling to embrace AI, time might be running out. In her article 3 AI-driven strategies for retailers in 2019, Giselle Abramovich claims “Personalization is table stakes for today’s retailers, who are increasingly competing to be relevant in the hearts and minds of shoppers.” Abramovich believes personalization might soon be a necessary condition for strong customer relationship management.
There is, of course, more than one way to win Olympic gold. Sometimes, recognizing your weaknesses and figuring out how to negate them against your competitors' strengths is the best path to victory. As the great Chinese general Sun Tzu said, “He will win who knows when to fight and when not to fight.” This was a strategy that Australia speedskater Steven Bradbury put to perfection at the 2002 Salt Lake City Olympic games.
According to Dr. Mussaad M. Al-Razouki, “Doing a Bradbury” has entered the Aussie lexicon to mean “to become the unlikely winner of a contest.” In his article Carpe Diem: Entrepreneurial Lessons From Olympian Steven Bradbury, Dr. Al-Razouki explains Bradbury knew he was the second-oldest competitor in the field and he had the slowest times, so he made a calculated decision — cruise behind his opponents and hope they crash.
Bradbury’s idea actually made a lot of sense for his situation and he implemented it successfully in both his semi-final and final races. “His reasoning was that risk-taking by the favorites could cause a collision due to a racing incident, and if two or more skaters fell, the remaining three would all get to the next round (or get medals in the final) and that since he was slower than his opponents, trying to challenge them directly would only increase his own chances of falling,” says Dr. Al-Razouki. As the races panned out, Bradbury had not one but two strokes of good luck when a crash occurred in both the semi-final and the final race. In the gold medal race, all four of his competitors crashed in the last corner because, as Bradley predicted, his competitors were aggressively jostling for gold. Bradley was about 15 meters behind the leaders, with only 50 meters to go, when the crash occurred. Bradley easily avoided the pile-up, then gracefully glided into Olympic history in one of the most remarkable finals ever skated. Second place was a comical keystone cops worthy slippery slide of skaters who looked more like amateurs taking to the ice for the first time than the Olympics gold medalists they were.
“Bradbury later admitted that he never expected all of his opponents to fall, but added that he felt that the other four racers were under extreme pressure and might have over-attacked and taken too many risks,” explains Dr. Al-Razouki. Bradbury thought each racer had their own unique desire to not settle for anything less than gold and that, as a result, would make them race overly aggressively and potentially knock each other out, which they did. “‘He will win who knows when to skate and when not to skate,” obviously, as well, to paraphrase Sun Tzu.
Bradbury’s strategy was spot on. For an athlete who had competed at several Olympic games for a country that never did well in the winter Olympics, Bradbury fits the old adage that says, “the harder you work, the luckier you get,” as Dr. Al-Razouki contends. However, Bradbury also proved without a doubt that being goal-directed helps you reach gold. AI might not help businesses win Olympic gold, but it does share a lot in common with athletes who do, even if some of its methods, like gold medal-winning speedskating runs and unsupervised learning, aren’t always the most understandable or even follow the most predictable of paths.