Today’s guest columnist is Tal Brown, founder of the human performance AI platform Zone7.
“Generative AI” is the topic du jour. Catalyzed by mainstream coverage of ChatGPT, “AI artists” and the White House’s AI Ethics Bill of Rights proposal, this content creation strand of artificial intelligence has captured imaginations, including those of professionals inside the sports economy.
Many businesses already use AI’s power to inform intelligent decision-making and gain competitive advantage. Machine learning algorithms can detect patterns, risks and anomalies more effectively than resource-intensive manual analysis.
I’m seeing firsthand how AI adoption, when done right, can fully leverage the full array of physiological, skill, strength and recovery data that athletes, leagues and teams curate. At Zone7, we identify hidden patterns across this data that precede injuries or periods of elevated performance. Our technology converts these insights into proactive training recommendations, which inform coaching decisions and help athletes maintain peak conditioning.
AI’s potential in athlete health and performance management is enormous, as are the financial implications for a sports organization. For example, we calculated the projected salaries for EPL players in the 2020/21 season and cross-analyzed them with injury incidence in the same period to determine the salary-value loss incurred by injuries across the league. In total, approximately $240M worth of “player days” were lost across 20 teams. Theoretically, a reduction of just 5% in injury rates would have saved around $7.5M in player wages. Many other benefits can be derived from keeping athletes healthy, including their impact on prize money, merchandise sales and sponsorships.
There are, however, two principles that sports organizations should recognize before leaping into the AI market. First, not all AI solutions are the same. The ones that inform human-decision making (as opposed to replacing humans) are more efficient and lucrative. Second, there must be technology and human foundations for data science put in place before adopting an AI solution, or else its benefits will never be fully realized.
This foundation should include the following elements:
Data Storage and Visualization Tools
These provide a snapshot of a department or organization’s present and past condition. They rely on raw data and are descriptive. Typically, coaches interpret these findings manually and structure their training regimes accordingly.
Conversely, AI solutions are prescriptive. They go one step further by using raw data and sophisticated algorithms to forecast future scenarios and create recommendations for desired outcomes.
Data Integrity and Human Input
You need a great deal of data to fuel an AI solution. The model needs to “learn” from many existing examples before it can be applied to live data and produce reliable outcomes.
To enable this learning process, organizations should store an inventory of datasets and ensure it is continuously refined and standardized. Where possible, a dedicated data manager should be on hand prior to AI adoption, to oversee data quality assurance and determine the best KPIs for measuring an AI solution’s impact.
This framework’s design has a direct impact on the ROI of the data being collected, the wearable devices collecting them, and any AI solution enlisted to extract important insights. The upfront investment required to get this right should be treated as an appreciating asset. It will pay dividends over time.
A strong AI solution is not “cookie-cutter.” It should be customized to the needs and preferences of a client’s sport, environment, roster and staff—including a coach’s preferred level of risk tolerance at any moment during a season, which will impact the frequency of insights and severity of alerts.
At Zone7, we frequently interact with client practitioners (coaches, physios, sports scientists, etc.); these practitioners must be an active part of calibrating algorithms and influencing how they work in their environment.
Robust Evaluation and Privacy Protection
The nascent AI solutions market has its share of credible and less credible offerings. Buyers need to conduct their own due diligence and test the mettle of solution providers with a thorough evaluation process.
Zone7, for example, tasks itself with providing custom validation studies for each prospective client. This includes an analysis of the client’s historical data, which verifies our AI’s ability to spot indicators of heightened injury risk, uncover other hidden insights such as the counterintuitive fitness parameters driving risk (e.g., not enough forward sprinting or lateral movement), and suggest modifications to on-field or off-field routines that can balance peak performance and fatigue.
No substantive discussion of AI adoption for an enterprise is complete without touching on privacy. Designers, developers and adopters of AI tools should secure permission to collect, process and analyze certain datasets. The most progressive leagues and teams are already including this language within individual contracts and collective bargaining agreements.
AI solutions can be game-changers, but only if they’re on the right playing field. An organization must determine whether it has a data science foundation in place to fully embrace AI, and ask itself the following questions:
- Does the department/organization have adequate datasets and/or data collection tools?
- Is there an existing architecture in place that newly adopted AI tools can integrate with?
- Would AI adoption improve ROI on other data collection and storage tools?
- Would AI adoption allow employees to shift their focus to higher-priority tasks?
- Is there research around an AI solution’s success for your desired application?
- Have you considered the primary and secondary costs of AI adoption, such as licensing, training and maintenance?
As “smart” as AI can be, it can’t work miracles. Human decision-makers must position the technology for success. The datasets a league or team collects, and the ontology with which those are handled, should determine which AI solution is adopted—rather than whether an algorithm is popular or “state-of-the-art.”
Tal Brown is co-founder and CEO of Zone7, an AI-based athlete health and performance platform used by organizations globally across elite soccer, basketball and American football leagues. His 20-year career in data science includes developing AI applications for Salesforce.
