Artificial Intelligence in Personalized Fitness Gets Smarter, For Real
Advances in AI mean apps, hardware, and biohacking companies will deliver exercise plans made just for you.
Long gone are the days when a personal training session required you to go to an actual gym to meet with an actual personal trainer. Today, there are myriad platforms, tools, and services that allow for personalization of your workout plan—and they’re only getting smarter.
Since Apple Watch launched in 2015 and began nudging us to meet new self-imposed “activity goals” (aka “close our rings”), digital platforms and tools have increasingly informed how we work out. Take Tonal, which launched in 2018; the cable-based weight system automatically provides you with the ideal amount of resistance to achieve your workout goals based on your individual strength. (The company saw 800 percent growth over the first year of the pandemic, and boasts a member “churn” rate—meaning the percentage of members who leave the program—of less than one percent). And the Future fitness app, which has raised more than $110 million over the last six years, is an app-based service that connects users with a personal trainer who provides customized workout plans—an option that used to only be available with a pricey gym membership.
Many more examples underscore the reality that now, thanks to recent advancements in artificial intelligence (AI) technology, the fitness world is doubling down on personalized fitness—creating bespoke plans, user accountability, motivation, and even a community that lives right in your smartphone, connected fitness product, or subscription service.
You can specifically thank generative AI, which is a type of artificial intelligence that can create new text, images, or other content based on raw data, for many breakthroughs in AI personal training. These programs can now process data from a variety of sources—whether that’s all of the articles on the internet about the best methods for building strength, the daily fitness habits of a large database of users, a customer’s personal genetic predispositions, or otherwise—and use that data to create fitness recommendations in a multitude of easily comprehensible formats that feel like a real person created them. (Even though, for the record, those recs are coming from a very sophisticated bot.) For the consumer, that might mean easier access to more effective, goal-oriented workouts that people are more likely to actually do.
“We are going to see a real implementation of AI powering personalization in 2024 because it's just gotten so inexpensive,” says Abby Levy, managing partner and founder of Primetime Partners (which has invested in Bold, a personalized fitness platform for seniors).
Cort Post, a principal investor at boutique sports, fitness, and gaming venture capital firm Courtside VC, is currently seeing a lot of beta testing of AI-driven fitness products. “We saw in the first half of [2023] the AI hype where everybody was raising money. And so the companies that could grab large rounds of funding are probably just now getting out in the market.” Take OpenAI, which runs the generative AI chatbot ChatGPT. The company received a $10 billion investment from Microsoft in January (although the specifics of that partnership are a bit in flux after a recent shake-up on the board), and is already partnering with multiple fitness companies, like Whoop (a fitness- and sleep- tracking wearable brand) and Tempo (a home gym company), to release generative-AI-powered fitness features this year and beyond.
However, the generative AI boom is only part of the personalization equation. The ability to bring together multiple data sources—including large datasets of user behavior and wearables that track activity, sleep, stress, nutrition, blood biomarkers, and more—to provide those “smart” customizations has also poised existing players in the smart fitness industry for personalization breakthroughs. Now that hardware-focused businesses—which collect intel on users’ strength, form, heart rate, habits, and more with each workout—have been around for a few years, they have more data than ever about how people are using their products to inform increasingly personalized programming and equipment. This data is then bolstered by more powerful algorithms to analyze those learnings.
Forging ahead on this front is Whoop, which released an OpenAI-powered health coach in its app in September. The health coach develops fitness (and rest) plans for users based on the questions they ask of the coach (like, “Can you make me a training plan for a 5K?”). The resulting workout programs draw on factors that Whoop measures, like users’ stress, sleep, strain, and recovery scores. Post, who has been following developments in AI coaching, says Whoop is the first to deliver on the ability to combine generative AI with granular biometric data like heart rate variability (or HRV) to create actionable, personalized plans.
"Now Whoop Coach can say, ‘Okay, not only is this what your recovery is today, [but] these are the reasons why your recovery is the way it is, and here's what you should do if you want a better recovery tomorrow and into the future.’”
Jaime Waydo, chief technology officer, Whoop
Jaime Waydo, chief technology officer at Whoop, sees the Whoop Coach as a natural extension of the brand’s mission. “What Whoop has worked on for years is taking complicated data and giving you something that's simple and actionable,” she says. “Whoop Coach takes it a step further because now Whoop Coach can say, ‘Okay, not only is this what your recovery is today, [but] these are the reasons why your recovery is the way it is, and here's what you should do if you want a better recovery tomorrow and into the future.’”
The lowest-lift consumer entry points into personalized fitness are free or subscription-based apps that create custom workout plans for users based on their goals, fitness level, available equipment, schedule, and other data points. Some, like AGIT and the JRNY workout app, use AI in computer vision applications (basically, using algorithms to analyze images and video) to offer corrections on a user’s form in real time or generate plans based on pictures of your equipment that you take and feed to the AI. Some people have even figured out how to plan ChatGPT workouts using just the ChatGPT interface. However, experts warn there are limitations to relying on generalized chatbots like ChatGPT for personalization.
“You can get amazing results that seem incredible, but it's not necessarily something that has been crafted with intention to make sure that you're safe and…specific to you,” says Justin Bingham, the chief technical officer of fitness training plan app FitnessAI.
FitnessAI, which costs $90 per year, has been making progressive overload weight training-based workout plans since 2019. In the new year, the company is debuting multiple apps that cater to the needs of different populations, such as one for powerlifters and another geared toward women over 65. Embedded in each app is a generative AI-enabled chatbot that will act like a personal-trainer surrogate, allowing users to ask for real-time adjustments and advice. Bingham says FitnessAI’s AI coach is more reliable and useful than getting advice from ChatGPT or a newer app, since it’s trained on the company’s data about how real-life users work out, and takes your own habits and metrics into account. FitnessAI has also acquired the apps of two fitness personalities, Althi by Linn Lowes (who has more than 3 million Instagram followers) and The Sculpt You by Katrina Wright. Fitness AI plans to develop text and animated image AI personas for the trainers, akin to the celebrity avatar technology released by Meta this fall, so that users can feel they’re being personally trained by someone to whom they have a connection.
Meanwhile, Virtuagym, which licenses fitness software to gyms and studios, launched an AI coach this year that lets people ask questions and make changes to their plans. In 2024, it is expanding its smart coaching capabilities to include nutrition plans, which means gym-goers will gain access to AI-driven personalized health and fitness for the price of their existing gym or studio membership. Another app company, Predictive Fit, uses your historical training data, as well as your raw DNA via 23andMe or Ancestry.com results, to recommend training plans for various sports. In 2023, the company also launched an AI-powered running coach called RunDot, which delivers insights on pace, training techniques, and more—costing between $13 and $149 per month depending on the level and frequency of insight you seek. In 2024, it will debut a similar program for cycling, called VeloDot.
Personalized fitness options are also becoming increasingly accessible—available right in your phone or smartwatch’s operating system—no extra apps required. With iOS 17 (launched on Apple phones in October), Apple Fitness+ users now have the option to create custom fitness plans: a weekly workout plan customized to your schedule, goals, and preferences is delivered straight to your phone. And with the Google Pixel Watch 2 (also launched in October), you can tap into the Pace Trainer feature, which provides real-time feedback on your pace (such as a direction to slow down or pick it up) when you’re training for a race or just running around the neighborhood.
Getting people to engage with these apps (and thus stick with their fitness goals) through the language or timing of notifications is another way some fitness apps are becoming more personalized. “We think a lot about building healthy habits and habit formation, and technology really knows how to use data [to understand] when's the right time to remind someone, what kinds of reminders, what kinds of things we can do to make a product stickier,” says Amanda Rees, co-founder of healthy aging fitness platform Bold. For example, the platform takes into account whether you're more likely to engage with it after an email versus a push notification, or a reminder at night versus during the morning—or, if reminders cause you to disengage altogether, it will adjust accordingly. The platform doesn't deliver the same messages in the same way to all users.
That’s not where the personalization ends with Bold, which aims to bring the “movement is medicine” philosophy to seniors for free through their insurance plans. New users complete a series of one-minute fitness assessments, such as how many times over 30 seconds you can stand up from sitting on a chair, as well as answer questions about fitness goals and health history, before Bold delivers custom workout plans on the platform with videos led by human instructors. In September, the company raised a $17 million series A funding round, and Rees says Bold is projected to be available to 12 million people next year, up from 10 million in 2023. That’s thanks to partnerships with insurance providers like United Healthcare where more insurees than ever before are opting into supplementary plans that grant access to preventative health care like Bold.
"We are looking at all your biometrics, [and] it's feeding into the workouts."
Tempo co-founder and CTO Moawia Eldeeb
Some companies in the personalized fitness space have a hardware component. This means users pay an upfront cost for equipment, which can be anywhere from a few hundred to several thousand dollars, as well as a monthly subscription that can range from the low double to triple digits. Take the aforementioned Tempo, a home gym company that provides personalized workout plans connected to weights and sensors that deliver form feedback. This year, it launched body composition scanning using computer vision on smartphones, allowing users to see how their muscle mass was changing and responding to training programs. It plans to unveil its own biometric data and AI-powered programming later this month or in early 2024. It will work by combining measurements captured with its own equipment (like body composition changes and real-time workout feedback) with data from wearables (like how well a person slept the night before). After synthesizing that information, Tempo will then suggest workout plans both for long-term practice and for making adjustments in the moment.
“Now Tempo has an AI [model] that knows your body in and out,” says Tempo co-founder and CTO Moawia Eldeeb. “We are looking at all your biometrics, [and] it's feeding into the workouts. Then at the end of the month, you're doing a scan. We're like, ‘Oh, we expected it to change your arms a little bit more than we thought. Let's change the workout further.’” Tempo is also working directly with OpenAI to deliver another AI-driven product, which it was not ready to share further details about, later in 2024.
Personalized fitness isn’t only for those working out from home, though. A new “smart boutique fitness studio” called Lumin opened its flagship location in Texas in September. With a space covered in LED screens, exercisers can choose an AI avatar (basically, a coach animated and powered solely through tech) that will coach them through workouts via their earbuds and as displayed on the LED screens. Cameras covering the whole space deliver form feedback directly to users’ smartphones. Gamification of the experience encourages streaks and personal records (PRs), while customized soundtracks sync to users’ music preferences and paces.
“Lumin co-founder Brandon Bean, the former CEO of Gold’s Gym, sees Lumin as a marriage of the recent boutique fitness and AI booms.
Photo: Lumin
Lumin co-founder Brandon Bean, the former CEO of Gold’s Gym, sees Lumin as a marriage of the recent boutique fitness and AI booms. “The consumer wants this group environment, but they also want to have a personalized experience,” he says. “We're trying to figure out a way to leverage technology to make that happen.” Bean says Lumin has yet to start marketing its studio, but the company already has “over 100” franchise leads across the globe, largely in the U.S. and Middle East.
The high-end options for personalized fitness often fall under the categories of biohacking, preventative medicine, and longevity. These offerings combine fitness with health and medicine in the form of lifestyle recommendations based on a person’s DNA and biomarker testing, as well as exercise, sleep, and nutrition tracking, and sometimes incorporate insights from continuous glucose monitors. These subscriptions can cost hundreds of dollars per month since they include regular blood (or even sometimes fecal matter or urine) testing.
One such medical-testing-powered fitness, nutrition, and lifestyle coaching company founded by self-help pioneer Tony Robbins, Lifeforce, raised a $12 million series A this year. It reportedly plans to use the investments to develop data tools that make personalization even more efficient. InsideTracker, which makes a host of fitness and lifestyle recommendations based on DNA and blood testing and fitness assessments like grip strength, plans to release a generative AI chatbot in 2024 that can answer questions about how you’re feeling and make recommendations, scan pictures of meals to detect and calculate macronutrient profiles, and provide “anomaly detection”—which will recognize when one of your biodata points is irregular, and give you insights about why that might be. In combination with gathering users’ wearable device data and blood test samples, health-coaching company BellSant does regular strength assessments of its users to determine how to optimize training for longevity markers like VO2 max. The company is experimenting with how to integrate lifestyle changes, such as nudging a user to turn a meeting into something they take while on a walk, to bridge the gap between making recommendations and implementing habit formation.
Investors and industry experts see these companies combining biometric, biological, and behavioral data as the ultimate direction in which the personalized fitness industry is heading. “Is that going to become the new normal, that we have this human performance dashboard [we look at], the way we open our inbox and email?” asks Levy. “I think that will ultimately become much more common.”
