Digital fitness: The emergence of AI-driven training platforms

triathletes-entering-water

When I signed up for my first Half Ironman two years ago, my approach to training was to go on Amazon.com and order a book that included three options for training plans based on available training time each week. I picked the intermediate plan averaging 10 hours a week, and over the course of 16 weeks completed the training with decent, but not spectacular results in the race. After signing up for another Half Ironman coming up in April, I decided it was time to revisit my approach to training to see if I could increase efficiency to improve results.

Historically, training plans for endurance sports have come in two forms, each with respective pros and cons:

1. Standard plans structured around hours per week or weeks remaining to prepare for an event. Just Google “4 month marathon training plan” and there is no shortage of standard plans. The benefit of these is they are either inexpensive or free. The drawback is there is limited customization in response to performance against the initial plan or in tailoring to specific goals.

2. Personal coaches offering varying levels of in-person or phone consultation. The benefit of a coach is they can optimize your plan against your goals and current training level, and adjust based on performance against the plan. The downside is that coaching is expensive, ranging from $200 to $400+ per month; putting it outside the budget for many individuals, especially at the recreational level.

A third option has now started to emerge: AI and big data-driven training platforms. This option combines the optimization of personal coaching with a significantly lower cost structure. The platform I ultimately decided to use, TriDot, starts at $29/month. These platforms allow the creation of an initial training plan that incorporates the individual’s fitness starting point, the time available each week for training and the particular event the individual is training for. Throughout the training process, the program then incorporates data from IoT devices on the individual’s training performance to modify the plan.

Traditionally, like healthcare and education, coaching is a labor-intensive business, and this both drives cost and limits ability to scale. If I open my own coaching business, while there is some overhead that will scale as I get more clients, in general, my tenth client will require as much of my time (and labor cost) as my first client. Additionally, quality becomes an issue as I need to hire additional coaches: How do I ensure their clients get the same quality of service as my initial clients?

On the other hand, what if I could translate my experience into an algorithm? If I can then build this algorithm into a platform that collects inputs and provides outputs, this revolutionizes my business model. Now my tenth client is significantly cheaper than my first. That additional client is just costing me, at most, some additional storage space, compute power and, potentially, service desk tickets — all significantly cheaper than a dedicated labor resource. I am also able to ensure that my ten thousandth customer has the identical quality experience as my first customer.

The first challenge that must be overcome is that AI needs data — algorithms drive off objective metrics. This seems obvious, but it doesn’t make it easy. A coach might alter a plan if an athlete feels overtrained, has a “bad” workout or feels a pace is too fast, but AI can’t run on “feelings.” We must be able to somehow translate these feelings or thoughts into quantifiable data. Fortunately it is possible do this — to quantify effort and recovery for optimized training through the use of IoT and online data collection platforms. But for now, please excuse me while I go hop on my bike. It’s a perfect summer morning to hit the trails.


Nicole Zenel headshotNicole Zenel works for DXC Technology as the Director of Global Deal Governance, Front Door and the Sales Information Center. In her spare time, she is an avid cyclist, triathlete and mountain climber. She enjoys learning and writing about how the latest IoT, big data and other technologies can be applied to these and other fitness pursuits. @NC_Zenel

Comments

  1. Love it Nicole….

    Like

  2. Great article. As an “older” MMA fighter, customization of workouts (intensity, frequency, activity), recovery, diet, and supplementation all go into the optimize plan. As you point out, “feelings” are key (I don’t spar unless I feel great as the outcome can be painful). Somehow this will need to be incorporate into any AI training program built for amateur athletes who train like pros. Looking forward to better apps in the future.

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: