An Elite Fitness Trainer’s take on trends in coaching tech

After 6 years and 4 courses (Personal Trainer course followed by specializations in Fitness Nutrition, Exercise Therapy, and Corrective Exercise), I’ve re-certified as an Elite Trainer!

During that time, my passion has been making strength training inclusive and accessible to everyone, especially women and people from other marginalized communities. I’ve welcomed scores of people into the weight room for the first time, including coaching several total beginners to medal in their first powerlifting meets. I’ve led my colleagues in weekly workouts and also taught “Googler-to-Googler” Intro to Powerlifting classes at a company gym. 

While personal training has never been my full-time job, I have coached for recreation and competition, in person and virtually, of various ages, in single and group sessions, and with clients of all experience levels. Here are my hot takes on two trends in fitness coaching tech.

AI Fitness Coach apps need to get more perceptive to safely train all users

First, dozens of new “AI fitness coach” apps have surfaced in recent years – even prior to Covid-19 (some examples). These apps use computer vision (e.g. through your laptop camera) and machine learning to understand how the user is moving and prescribe workouts. The hypothesis is that they can scale coaching value to more consumers by replacing human trainers.

Backing up a bit, when I’m crafting a program or session for new client(s), I’m considering many factors, including:

• What equipment is available and accessible?
• Is there anything the client’s doctor has specifically told them to do or not do?
• How familiar is my client with specific exercises?
• What kinds of exercises do they like or dislike?
• How flexible is my client?
• What muscle imbalances exist with my client?
• What is my client’s body awareness?

These last 3 questions are doozies, because they aren’t commonly considered or easily assessed. For example, body awareness (also known by its fancy name proprioception) is the sense of self-movement and knowing your body’s current position. When I’m coaching, I assess body awareness by watching the client’s performance in various movements. If someone has high body awareness, I can safely teach more new movements faster, suggest a larger breadth of independent workout plans, and use more types of cues – or vice versa.

Unfortunately, pose tracking-based AI coaches aren’t anywhere close to precise enough to evaluate body awareness, flexibility, muscle imbalances, or more. But because these apps are branded as AI fitness coaches, users may be misled to believe that they’re receiving attention equivalent to an actual human coach. In reality, the apps are like a lackadaisical human coach: it thinks you should do this exercise, but it doesn’t really know if you can do it without injury. 

Several things should change:

• Apps should set the right expectations. Claiming an AI coach will count reps or distinguish if you’re doing a squat versus a push up? Go ahead! Claiming the AI coach will catch all form mistakes, even something like slipping out of a neutral spine (the safe spine position)? Probably disingenuous, and lulls people into a false sense of safety.
• Apps should skew toward safer exercises. More people would be able to do a squat safely than a jumping lunge. Smart programming can help avoid the most dangerous exercises for less prepared clients.
• Build more precise computer vision models for known problem areas. There are some specific common, important, and hard-to-detect form mistakes across exercises, such as not keeping a neutral spine or shrugging your shoulders. Models trained at the granularity of the whole body can’t detect these nuanced errors, but models could be trained specifically around these expected mistakes.

So AI fitness coaches have a way to go to replace in-person trainers. What about when a human is coaching, but over video chat?

Live coaching over video chat has a set of unique challenges

When I’m leading a class in person, I’m constantly moving around the floor, examining clients’ form in 360-degree view and giving individual encouragement and cues. I’m asking myself questions like:

• Can everyone see and hear me when I’m explaining the plan or new movements?
• Does everyone know what they should be doing, and are they doing it?
• Are they doing it safely, with correct form and the right weights?
• Is everyone having fun?
• How much time left until we’re changing to the next set or exercise?

Over video chat, everything changes:

• I’ve got to fit my whole body on the screen, and present different angles to the class.
• It’s hard to check everybody’s form, since I can only see people from one angle, the camera view is sometimes wrong, and they’re condensed to a tiny 3-inch human on my screen. I need to be both far enough away to remain in view, while close enough to evaluate a tiny pixelated person’s form.
• Building a fun, upbeat energy is harder — music is difficult to synchronize and share. Conversation doesn’t flow naturally, and interacting requires speaking up to the whole group, whereas in person they’d be able to joke or speak with the people beside them.

There’s an opportunity for a video chat workout platform to do more: play synchronized music that auto-quiets when the coach is talking; sequentially spotlight clients on the screen for easier form evaluation; auto-correct clients when they wander off-camera; recapture that “workout buddy” feel by highlighting one buddy who they’re working out “with” on screen and can talk privately to; sync phone and laptop cameras to show multiple angles of the coach; and more. 

Fitness coaching is moving more online, driven by Covid-19 and by sheer convenience. There’s opportunity to make fitness more accessible to more people, and video chat platforms and AI coaches have key parts in this distributed-fitness future. But we have an ethical obligation to build these trends safely, with users aware of what they’re getting into, and trainers empowered to keep clients safe.

Stay tuned for pt. 2, where I’ll discuss other aspects of fitness coaching tech, like 1-on-1 remote training (e.g. Future Fit) and on-demand courses.