November 2026
By Al Noone · Last updated November 13, 2026
Imagine you work for a large technology company, the kind that builds AI products and studies everything its users do. One morning an email arrives from HR announcing a new benefit: a free AI personal trainer, built in-house, available to every employee. The company, the email says, cares about your wellness.
Sure it does. (At least, it cares about your wellness to the extent that your wellness affects its bottom line.) But you know how your employer makes money. It collects data on user behavior and turns that into predictive profiles, used to sell targeted advertising. Now, besides tracking you at work, your employer would learn when you work out, how often you skip, what hurts, and how you're sleeping. Even if you wanted a coach who could do all that, you might rather pay for one yourself than accept the free service from your employer.
Then a second email arrives: employees who opt in get a discount on their health insurance premium. You're not forced into it, but declining to use the wellness benefit costs you money. The program doesn't feel so voluntary.
HR should put itself in employees' shoes and expect employees to resist an offer like this. Well-intentioned though such a wellness benefit may be, the structure strains the credibility of promises that employee exercise data wouldn't be used for purposes beyond simply helping employees with their wellness goals. To fix the problem, an independent party (not the employer, not the health insurance company) can run the wellness program and hold the data. That way, employees don't have to take the employer's word for anything.
Many employers offer an Employee Assistance Program (EAP), a confidential counseling service for stress, grief, addiction, money trouble, and family problems. Most hire an outside vendor to run it rather than staffing it themselves.
Employers could offer an in-house EAP. They have HR departments, health plans, and budgets. They outsource it anyway because the program only works if employees use it, and employees won't confide in a counselor who reports to the people who decide their raises. The vendor's independence is a necessary feature if an EAP is to be used.
A personal trainer learns things about you that you might not share at work. The old back injury. The weird clicking in your ankle. Feelings of lassitude. Whatever. A trainer who knows these things coaches better. An employer who knows them could use the information, deliberately or not, in decisions about promotions, layoffs, or insurance.
That's the trouble with the in-house trainer in the fictional-yet-plausible scenario. The employer might not misuse data, but employees can't be sure, and a trainer you can't be honest with won't be as good. The EAP model solves this for fitness the same way it does for counseling. The employer sponsors the program, an independent provider delivers the service, and the employer sees aggregate numbers, like how many seats are in use, rather than anything about a particular person.
The same reasoning applies outside the workplace. You could get coaching by talking to a general AI assistant directly, and we've written about that choice in Why use Xenos Fit if you already use AI. The short version: there, your training details sit alongside everything else tied to your account.
Economists call this a credible commitment. A promise is so much more believable when the promiser is unable to break it. An employer holding your health data can promise never to look, but they might change their mind. An employer that never holds the data doesn't need to be trusted on that point at all. Thomas Schelling, who studied this kind of strategic self-binding, showed that giving up your own options can be the most persuasive move available.
Legal scholars have a related idea. Jack Balkin has argued that companies holding our personal information should be treated as information fiduciaries, owing us duties of care and loyalty the way doctors and lawyers do. And the philosopher Helen Nissenbaum's framework of contextual integrity explains why the in-house trainer feels wrong before anything bad happens. Information shared for one purpose, like coaching, carries expectations about where it goes next. Training data going into a personnel file breaks those expectations.
Involving a third party doesn't remove the requirement for trust; it transfers the trust from the employer to the third-party provider, for a scope of service and data handling. Within that scope, the third party should deserve more trust than the employer.
This isn't a new idea either. Decades ago, John Hagel and Jeffrey Rayport predicted the rise of “infomediaries,” companies that would manage people's personal data on their behalf. Few took hold, and every intermediary faces the same temptation the model was meant to prevent: monetizing what it knows about the people it serves. So the question to ask of any third party is whether its incentives are structurally different, not just its intentions. Here's how Xenos Fit answers that question.
Your data is never used for advertising or shared with anyone who would use it that way. Nothing you tell your personal trainer affects what ads you're shown. It isn't sold to data brokers either, so it never gets merged with your purchases, location, and browsing into a profile that follows you from company to company. Individuals or their employers pay for the service, and nobody else does. There's no advertiser or data buyer whose interests compete with yours.
We store as little as we can. Data that doesn't exist can't be breached, subpoenaed, or repurposed later.
You don't have to tell us who you are; we don't need to know. We encourage alias email addresses and pseudonyms.
Employers see aggregate numbers, not details on individual employees. A sponsoring employer gets a monthly report of group totals, like how many seats are in use and what share of people trained each week, but never who is training, how often any one person trains, or for what. Any count below five people is reported as “fewer than five,” so a small group can't be picked apart.
Xenos Fit runs on AI models from other companies, so those providers do process your conversations. They process them as Xenos Fit's traffic, though, not as part of a personal account of yours with them.
AI is making surveillance cheaper and more persuasive. The same tools that can coach you can profile you, predict you, and sell to you with unnerving precision. But you can have the benefits of AI while skipping the creepy side effects by keeping your data split up, so no single company sees the full picture, and choosing providers whose incentives line up with yours.
Further reading