- Virtual fitness platform Future shut down its AI personal training beta on June 19, 2026, citing member feedback and a Les Mills report finding that roughly nine in ten consumers globally prefer a human coach over an AI alternative.
- AI fitness apps including Fitbod and Virtuagym offer adaptive programming, real-time feedback and plateau detection that human coaches cannot replicate at scale across large user bases.
- Future’s human-only memberships run from $129 to $399 monthly, a price point that suggests the accountability and behavioural depth of human coaching commands a durable premium that AI tools have not yet displaced.
Future just shut down its own AI coaching product. On June 19, 2026, the virtual fitness platform ended its AI personal training beta and returned fully to human coaches, citing member feedback and a Les Mills report finding that roughly nine in ten consumers globally prefer a human trainer over an algorithm. The decision is worth examining closely, because the question it raises matters well beyond one company’s pivot: where do AI fitness tools actually deliver, and where do they consistently fall short?
The Human Touch in an AI Era
According to Future, its members pushed back on the AI offering. The Les Mills 2026 Global Fitness Report, cited in the company’s announcement, found that roughly nine in ten consumers globally prefer guidance from a human coach over AI alternatives. That finding does not invalidate AI fitness tools, but it does set a clear ceiling on how far algorithmic coaching can go before users disengage.
AI’s Precision Edge: Adaptive Training Algorithms
Future’s exit from AI coaching does not reflect the broader market. Apps like Fitbod and Planfit have built their products around adaptive programming that is genuinely difficult for a human trainer to replicate at scale. Fitbod tracks muscle fatigue across sessions and adjusts exercise selection to avoid overworking recovering muscle groups. Planfit gradually increases workout difficulty based on user input, adding variety to prevent stagnation. Both apps draw on a user’s performance history, equipment availability and stated goals to adjust training load over time.
A human trainer can do the same thing, but not for hundreds of clients simultaneously and not with the same granularity of logged data. For users who train consistently and want a programme that adapts without requiring a coaching call, that kind of algorithmic personalisation has real practical value.
Beyond the Numbers: Multimodal Data and Real-Time Feedback
Some AI systems can classify exercise execution quality, including detecting breathing irregularities and asymmetric muscle effort. That kind of instant, specific technical feedback is useful, particularly for people training alone without access to a gym or a coach. The value compounds over time: the more sessions are logged, the more precisely the system can calibrate its recommendations.
The Rise of Generative AI in Personalisation
Generative AI is changing how fitness apps handle user interaction. Rather than navigating fixed menus or pre-set routines, users can engage with conversational interfaces that respond to real-time input, reported feelings and direct questions. The practical difference is responsiveness: a chat-based AI coach can explain why a workout is structured a certain way, adjust on the fly when a user flags fatigue and provide motivational prompts without requiring a human on the other end.
Whether that interactivity drives retention in practice depends on the implementation. The underlying claim, that conversational interfaces feel less prescriptive than static apps and therefore keep users engaged longer, is plausible on its face. How widely it holds across different user populations is harder to verify from public material. What the generative layer does clearly add is accessibility: it makes the outputs of adaptive programming and real-time feedback easier for non-expert users to interpret and act on.
Navigating the Plateau: AI’s Role in Overcoming Stagnation
Plateaus are where many fitness programmes break down. Progress stalls, the user assumes the programme has stopped working and either pushes harder or quits. AI apps approach this differently. Rather than waiting for a user to notice stagnation, they monitor trends across weeks of data and flag when progress has genuinely stopped rather than just fluctuated. By correlating training logs with sleep quality, recovery metrics and session completion rates, an AI can identify likely causes, whether that is insufficient sleep, too little variation in exercise selection or inadequate recovery between hard sessions, and suggest specific changes.
Virtuagym‘s AI Coach applies progressive overload principles automatically, adjusting weights and volume based on what the user logged in previous sessions. If the adjustment produces no measurable response within a set period, the system recommends further changes. For users who tend to repeat the same workouts indefinitely, that kind of structured intervention is more likely to produce results than willpower alone.
The Indispensable Human Element: Why Future Pivoted
The case for human coaches is not primarily about data. A human trainer can hear that a client sounds flat on a call and adjust the session on the spot. They can recognise that a missed workout signals something going wrong at home rather than simple laziness. They can build the kind of trust that makes a client honest about what they are actually eating or how much they are actually sleeping.
Evaluations of models like OpenAI‘s GPT-4 in exercise prescription contexts have found that while the model produces broadly safe and reasonable programmes, it tends toward excessive caution rather than training effectiveness and cannot reliably account for individual health conditions outside standard parameters. Future’s member feedback, as the company reported it, pointed to this gap directly: people wanted the accountability and adaptability of a human relationship, not just a programme that updated itself. That is not a technical gap AI is likely to close quickly, because it is not primarily a technical problem.
Economic Trajectories and Market Segmentation
The fitness technology market is splitting along two distinct lines. At one end, AI-driven platforms like Fitbod and Freeletics compete on scale, price accessibility and algorithmic efficiency, serving users who want data-driven programming without the cost or scheduling overhead of a human coach. At the other end, Future’s human-only memberships, priced from $129 to $399 monthly, demonstrate that a meaningful segment of users will pay a substantial premium for accountability, empathy and the kind of nuanced adaptability that sits outside a data model’s reach.
These two segments are not really competing with each other. AI excels at the quantitative side of fitness: tracking, adjusting, detecting patterns across large data sets. Human coaches remain the better option for addressing the psychological and behavioural components of long-term wellness. Regulatory pressure on data privacy and algorithmic fairness is also likely to increase compliance costs for AI providers, which may further sharpen the segmentation as trust and transparency become more visible purchase criteria. For enterprises building or procuring fitness benefits platforms, this split matters: the right tool depends on which problem you are actually trying to solve. For a broader look at how enterprises are navigating AI agent adoption more generally, the recent Sinch survey on AI agent rollbacks is instructive.
Ethical Considerations and Data Privacy
AI-powered fitness wearables and apps collect detailed personal health data, and the governance standards around that data vary considerably across the industry. The risk of algorithmic bias in these systems is real, and transparency about what data is collected, how long it is retained and who can access it is not a given. Companies that get ahead of this, through clear disclosures and auditable data practices, are better positioned as regulatory scrutiny increases.
What To Watch: The Hybrid Future of Fitness
The most likely outcome is a durable market split rather than one approach displacing the other. AI fitness tools will keep improving at the data-driven end: adaptive programming, real-time form feedback, plateau detection, recovery monitoring. These are areas where continuous data processing produces a genuine advantage over what a human coach can manage across a large client base. At the same time, platforms like Future are demonstrating that a meaningful segment of users will pay a significant premium for human coaching, particularly for anything touching on long-term behaviour change, complex health goals or accountability that cannot be manufactured algorithmically. Understanding which of those two things a user actually needs is the most practical takeaway from Future’s decision. Stay up to date with the latest AI developments at Auton AI News.
