Echelon Fitness is rolling out a generative artificial intelligence (GenAI) assistant for the users of its fitness equipment that adds a unique personalization capability: You can “talk” to a chatbot to make changes to your custom workout plan.
See too many 45-minute workouts in the schedule? Audibly tell the AI to swap those out with shorter regimens, just like one would speak to ChatGPT. Going on a business trip? Tell the AI to only schedule exercises that don’t require gym equipment for the days you’re out of town.
The launch comes at a time of increasing interest in AI-powered fitness capabilities, such as those offered by Peloton, Google Coach, Strava and others. While many fitness companies or apps use AI to customize workouts for the individual, most don’t offer a voice conversational AI.
The chatbot is part of the new Echelon Intelligence platform developed in collaboration with AWS. It uses generative AI to create 2-week custom fitness plans based on the user’s workout history, style and preferences. The chatbot, which is included in Echelon’s Premier plan, can make changes to the workout schedule just by telling it to do so.
Echelon aims to address a big challenge in fitness: Make sure users stay on their workout regimen whether they’re at home or in the gym.
“That was one of the things we wanted to do — bridge that gap between home and gym fitness,” said John Santo, co-founder and chief product and technology officer at Echelon, in an interview with PYMNTS.
“Everybody’s sort of hybrid now. You work at home a little bit. You go to the office,” he said. “Now, you go to the gym; you work out at home. So we wanted to bring that together, and we did.”
PYMNTS Intelligence data has also shown that consumers are looking for easy ways to multitask and get the most out of their time.
Read more: Fitness Industry Looks to AI to Bring in Consumers, Profits
Echelon makes treadmills, exercise bikes and other fitness equipment. It offers live or on-demand fitness classes in equipment with screens. Santo said Echelon serves over a million users globally and its equipment is found in 8,000 gyms, hotels, apartments and fitness studios.
Santo said Echelon used Amazon GenAI models to analyze members’ fitness assessments and goals. Amazon OpenSearch’s vector search can match users with classes and instructors based on their workout history.
AI also analyzes what factors made a user stick to working out and offers suggestions.
For example, if it notices that a user worked out in the same live-streamed class as another individual and they both belong to the same Echelon community group, the AI can suggest connecting to the other person as an accountability buddy to encourage each other to keep going.
“Discouragement is the thing that kills fitness,” Santo said.
Early testing indicates that GenAI customization could increase customer retention by at least 7%, which Santo said is a “meaningful” number.
See Also: Peloton Continues Comeback With Strong Subscription Metrics
Asked what made Echelon tackle GenAI when many companies still shy away from full deployment, Santo said the company has always been data-driven. Using GenAI that constantly learns from new data will help their customers hit their fitness goals, he said.
Key to deployment was working with AWS engineers who taught Echelon’s technical team how to maintain the generative AI system. Santo said his team learned to add content to the AI by themselves.
“If you start now and start small and get your team exposure to it, then they become more comfortable with it and it becomes a lot easier for them to adopt it and the resistance becomes less,” Santo said.
Generative AI also did not balloon Echelon’s IT costs. Santo said it added at most 5% more because of the way they structured the system.
Generative AI can be a variable cost. Any time a user prompts a chatbot and gets an answer, the amount of text sent and received carries a fee based on volume. But Santo said the system creates a 2-week schedule based on a user’s past data, and users only interact with the chatbot to make changes. This keeps the costs down since the chatbot is not writing essays.
“We’re not incurring a lot of LLM fees going back and forth all the time,” Santo said.
Related: Top Fitness Apps of 2024: Features Elevate User Experience and Health Goals
However, Echelon did have to prepare its data for AI. Santo said the company went through “a lot” of rounds of sanitizing its data and inputs that go into the AI model.
Santo noted that Echelon has been interested in using machine learning since 2019 but realized that the data within its organization and the structure of its content platform were not set up for it.
For example, “we redesigned our entire content library for better search. We put it on (AWS’) OpenSearch, where it’s more like a Google search,” Santo said. “It’s a lot more organized now.”
Their data-cleaning efforts led Echelon to generative AI. “It really was like a long journey to get where we are,” Santo said. “We always try to have a three-year ‘look-forward’ plan. … That really helped us.”
Sanitizing the data also helped reduce hallucinations. Santo said there typically isn’t a lot that occurs, since the chatbot isn’t asked about facts nor does it create its own content. But the AI can get dates and times wrong.
“One of the things LLMs really have trouble with is telling what time of day it is or what day it is,” Santo said. “Not very good at it.”
Looking ahead, Santo said Echelon plans to add live classes into personalized workout schedules within the next 30 days. The company is also working on a white-label app it can license to gyms that can generate workout plans based on specific gym layouts and available equipment.
“We’re really going to be pushing hard into the commercial space in the next few months,” he said.
