Why the Future of Work Is Self-Employed: AI, the Gig Economy & the End of the Corporate Job

Independent workers are projected to be 48.5 percent of the US workforce by late 2026 and a majority by 2027. AI-attributed layoffs grew 12x in two years. Median W-2 tenure is at its lowest since 2002. The corporate career is unbundling — and the people who built durable, in-person service businesses already crossed to the other side of the curve.

Jesse Ray Snyder

B.S. Exercise & Sport Science · NASM CES · May 2026

The conversation most trainers are having about their career right now is the wrong conversation. They are asking whether to leave their gym for an independent practice. They are weighing risk against stability. They are running scenarios where the variables are the same ones their parents used to choose between Sears and the post office in 1978.

The variables have changed. The labor market is in the middle of the largest structural unbundling since the industrial revolution, and the direction of the unbundling is unambiguous: away from W-2 corporate employment and toward independent, owner-operator work. This is not a forecast based on hopes. It is the actual measured trajectory of US labor data, AI capability development, and tax-and-payment infrastructure built over the last 15 years.

This article makes one argument, supported by Bureau of Labor Statistics data, IRS reporting infrastructure changes, Challenger Gray and Christmas layoff tracking, MBO Partners and Statista independent workforce projections, and the actual capability curve of generative AI: the future of work in the United States is majority self-employed, and the trainers who built documented in-home subscription practices have positioned themselves into one of the few work categories that AI cannot displace.

I am not making a prediction. The data has already moved. Most people have not looked up from their screens long enough to notice it.

The Trajectory Is Already Decided

Pull back from the daily news cycle and look at the labor data over a long horizon. The story is consistent.

In 2024, approximately 70 to 76 million Americans participated in some form of independent or freelance work. Estimates vary by source — the World Bank, the BLS contingent worker survey, MBO Partners, and Statista all use slightly different definitions — but the directional figure is consistent. Independent workers are projected to reach 86.5 million by 2027 per multiple industry forecasts. US freelancers are expected to represent approximately 48.5 percent of the total workforce by late 2026, approaching a historical majority forecast for 2027.

The income tier is also expanding. 4.7 million US independent workers earned over $100,000 in 2024, up from 3 million in 2020 per MBO Partners. That is a 57 percent increase in high-earning independents in four years. Independent work is no longer "supplemental income for people between real jobs." It is the work, for a meaningful and rapidly growing share of the American professional class.

Layer on the W-2 tenure data from the Bureau of Labor Statistics. Median tenure with a single employer in the United States is 3.9 years as of January 2024 — the lowest figure recorded since January 2002. Personal care and service occupations show 2.5 years. Workers ages 25 to 34 show 2.7 years. The "career employment" model that dominated mid-twentieth-century work has not been the actual median experience for at least two decades, and the median has been compressing further every year.

Both curves point the same direction. The independent workforce is expanding. W-2 tenure is contracting. There is no scenario consistent with the available data in which corporate employment dominates the work landscape in 2035. The line items are already on every credible projection. The only debate is the slope, not the direction.

What Is Actually Happening to Corporate Employment

The above is the long-horizon picture. The short-horizon picture is more pointed.

According to Challenger, Gray and Christmas, US employers announced approximately 1.17 million layoffs in 2025 — the highest annual figure since the 2020 pandemic. In the first two months of 2026 alone, technology firms announced approximately 32,000 layoffs. The pace has not slowed in the spring.

What is new about this cycle is the framing. Of the 1.17 million layoffs in 2025, approximately 55,000 were directly attributed to AI by the announcing companies — more than 12 times the number of AI-attributed layoffs just two years earlier. Major individual announcements during 2025 and early 2026:

1. Amazon: 14,000 corporate roles eliminated. The largest single layoff in company history. CEO Andy Jassy said the company expects to shrink its white-collar headcount further as AI agents are deployed.
2. Workday: 1,750 jobs (8.5 percent of workforce). Cited as a reallocation toward AI investments. A SaaS company laying off SaaS workers to fund AI tooling that displaces SaaS workers.
3. Block: nearly half of its 10,000-person workforce. CEO Jack Dorsey said AI had made many of those roles unnecessary and predicted that "within the next year, the majority of companies will reach the same conclusion."
4. Chegg: 45 percent of workforce. Cited as the "new realities of AI" and reduced traffic from Google.
5. CrowdStrike: ~500 positions. Cited explicitly as AI reshaping operations.
6. McKinsey: ~200 technology and professional support staff. The firm that built its business advising other companies on whom to lay off began the same process on its own analyst class.

These layoffs are concentrated in white-collar knowledge work — the categories that previously promised the highest career stability. Customer service, software development entry-level roles, marketing operations, finance and accounting, content writing, research and analysis. Wall Street banks plan to remove approximately 200,000 roles over the next three to five years, with the cuts concentrated in entry-level and back-office positions.

The honest read is that the AI capability curve is moving faster than the labor market's reabsorption rate. Historically, automation has been offset by the creation of new types of work — the "reinstatement effect." Recent academic research from Acemoglu, Restrepo, and Autor suggests that displacement is now intensifying while the creation of new work has not kept pace. The white-collar career model is undergoing a structural compression, and there is no credible argument that the compression reverses in the next decade.

The AI Layer: What Is Actually Being Automated

To predict what survives, you have to be specific about what AI is actually replacing. The cultural conversation has gotten this wrong in both directions. AI is not "replacing everything," and it is not "just a tool that helps everyone." It is doing something more specific.

Generative AI — large language models, image and video generation, code synthesis, voice cloning, agentic systems — is automating cognitive tasks that can be performed at a keyboard from a remote location. That is the specific category. Sub-tasks within that category include:

1. Pattern-matching from text input. Reading documents, summarizing meetings, drafting emails, translating between formats, answering customer questions from a knowledge base.
2. Generation of structured content. Marketing copy, code, financial models, exercise programs, meal plans, legal first drafts, design layouts.
3. Routine analysis. Reading reports, extracting data, comparing options, ranking alternatives, performing standardized calculations.
4. Asynchronous communication. Customer service chat, scheduling coordination, follow-up sequences, content moderation.

What AI is not automating, at any meaningful scale, and almost certainly will not be automating within the next decade for trainers' practical purposes:

1. Physical presence in a human's home or space. A robot capable of driving to a client's house, climbing the steps, entering the kitchen, setting up training equipment, demonstrating a hip-hinge while the client watches, and physically spotting a heavy press does not exist at any consumer price point. The general-purpose humanoid robot is a research project. It is not a product trainers compete with.
2. Real-time embodied movement correction. Watching a client's squat, feeling whether they are bracing, cueing the next adjustment, and adapting the program to what their nervous system did this morning is a multi-modal task that current AI systems perform poorly on even in lab conditions.
3. Trust-based relationship work. Clients hire trainers in their homes partly for the work and partly for the relationship. The accountability of someone showing up, the social structure of a scheduled session, the trust required to be vulnerable about your body and habits — these are not "features" AI can optimize. They are the substrate of the entire transaction.
4. Local-knowledge service judgment. Knowing your specific knee history, your work schedule, the fact that you travel to Phoenix every third week, your spouse's surgery recovery, your kid's basketball schedule that just shifted, and how all of that should reshape the next four weeks of programming — that integration of long-running context with judgment is exactly the type of work where AI has the worst leverage relative to a trusted human.

The framing I find most useful is: AI is automating the cognitive layer of work fastest, and the embodied/relational layer of work the slowest. A trainer who built their business around the cognitive layer (exercise programming as the differentiator, the workout-as-product) is structurally exposed. A trainer who built it around the embodied/relational layer (presence, accountability, hands-on correction, long client relationships) is structurally protected. I argued the specific version of this argument for trainers in a previous piece.

Why In-Person Service Work Is Structurally Protected

The protection in-person service work has against AI displacement is not motivational. It is mechanical. Three specific mechanisms make this category durable:

Mechanism 1: The robotics gap

For an AI agent to replace an in-home personal trainer, it would need to: leave a server, animate a physical body, navigate to a client's house, identify the right address, enter the home, recognize the human, set up training, demonstrate movements, spot loads safely, observe form, give haptic and verbal feedback, adjust based on real-time signals, document the session, and leave. Every one of those subtasks is a hard robotics problem. Several of them are currently unsolved in any product-quality form.

The cognitive layer of training — "what program should I do" — will be solved by AI quickly, and is largely solved already. The physical-presence layer of training will not be solved by AI on any reasonable horizon. The gap between these two timelines is the structural protection the in-home trainer is positioned inside.

Mechanism 2: The trust premium

People will pay a substantial premium for human-delivered services when the alternative is an AI version, in categories where trust matters. We are watching this premium increase, not decrease, as AI becomes pervasive. The categories with the largest trust premiums are: physical body work (training, massage, physical therapy), medical care, in-home services (cleaning, repair, childcare), and high-stakes professional advice (legal, financial, real estate). Personal training sits in the strongest of these categories — high-frequency, high-trust, in-home, body-related.

The AI version of "exercise programming" is a free chatbot. The human version is a trainer the client trusts in their kitchen at 6:30am. These are not substitutes. They are different products competing for different jobs-to-be-done.

Mechanism 3: The accountability mechanism

The reason most fitness products fail — apps, online programs, free PDFs, virtual coaching — is not bad programming. Programming is the easy part. The hard part is getting someone to actually train consistently for years, especially after the novelty wears off. The trainer in the home solves accountability with a mechanism that AI cannot replicate: a scheduled appointment with a specific human who is going to show up whether or not the client feels like training.

You can simulate this with an app. You cannot replicate it. The empirical evidence is the multi-billion-dollar app graveyard littered with high-quality fitness products that failed because they could not solve adherence. The mechanism for adherence is not a feature. It is a person.

The Three Tailwinds Compounding Behind Self-Employment

Beyond AI displacement of W-2 work, there are three independent infrastructure tailwinds that have made the shift to self-employment dramatically more accessible than at any previous point in history.

Tailwind 1: Payment infrastructure

Stripe launched in 2010. Square in 2009. The full modern stack — recurring subscription billing, ACH at marginal cost, instant payouts, dispute handling, fraud protection, mobile-first checkout — reached operational maturity around 2016. The infrastructure to run a recurring-revenue service business out of a single person's iPhone did not exist before this point. It now exists for under $0.30 per transaction.

I ran six years of Stripe subscription billing across Monterey Personal Training with zero chargebacks. That outcome is not possible in 2002. It is barely possible in 2010. It is mundane in 2026. Payment infrastructure has eliminated one of the largest operational barriers to running an independent service business at scale. The mechanics are documented in my pricing and billing breakdown.

Tailwind 2: Local discovery infrastructure

Google Business Profile, local SEO, Google Maps, and the Apple Business Connect ecosystem have made it possible for any single-operator business in any town to be findable by clients searching for a specific service in their specific zip code. This was not true in 2005. Local visibility used to require Yellow Pages spending, newspaper ads, and physical signage. It now requires a half-day of profile setup and a few months of accumulating reviews.

This means a single trainer in a mid-sized city can be the top organic result for "in-home personal trainer near me" with zero advertising spend, simply by claiming the profile and earning reviews. The independence math used to be gated by the cost of acquiring customers. That gate has effectively dissolved for local service businesses. I covered the practical playbook in my Google Business Profile article.

Tailwind 3: Tax and reporting infrastructure

The IRS has spent the last decade quietly modernizing the infrastructure for independent contractor work. The 1099-NEC form was reintroduced in 2020. The 1099-K reporting from third-party processors has been steadily expanding. The Schedule C tax treatment for sole proprietors and the qualified business income (QBI) deduction under Section 199A make the post-tax economics of self-employment materially more favorable than W-2 employment for most income tiers.

The downstream effect is that the infrastructure assumes self-employment is a major and growing share of the workforce, and is being built to accommodate that. The tax code is now structurally biased to make independent work more financially attractive than it has been at any prior point. That is not an accident. It is policy keeping pace with where the workforce is already going.

Why Personal Training Sits at the Center of This Convergence

Step back and stack the four forces:

It is hard to overstate how unusual this position is. Most work categories sit in exactly one of these zones. Some are protected from AI but lack operational infrastructure. Some have great infrastructure but are first in line for displacement. Some are both protected and well-infrastructured but require capital intensity (medical practices, restaurants) or licensing barriers (law, real estate) that filter most operators out.

Personal training is one of a small handful of work categories that is simultaneously: protected from AI displacement (physical presence and trust), supported by mature operational infrastructure (payments, local SEO, tax structure), accessible to single-operator entry (no large capital requirements), and growing in latent demand (aging population, sedentary white-collar workforce, rising health awareness). The convergence is structural, not lucky.

The trainers who recognize this are positioning into the next decade. The trainers who are still arguing about whether to leave their gym are positioning into the last one.

What This Means for Trainers Specifically

I want to be specific about what changes in this analysis and what does not.

If you are currently a W-2 trainer at a commercial gym, the trajectory of the gym-employment model in your specific location is almost certainly bad. The mechanism is straightforward: large commercial gyms operate on thin margins and are subject to the same cost-cutting pressures hitting every other multi-location business. As AI tools make it cheaper to operate without trainer staffing — or to use lower-skilled staff with AI program-design support — the economic incentive for gyms to maintain a large trainer payroll declines. The split that's currently 30/70 in the gym's favor compresses to 20/80, or the trainer position becomes part-time, or the gym shifts to a small-group model that requires fewer trainers per member.

Meanwhile, the in-home subscription trainer is operating in a category where AI cannot directly compete. The trainer's product is not "an exercise program." It is "a trusted human who shows up at your house and watches you train, every week, for years." That product is durable in a way that "programming" is not.

The strategic implications:

1. The trainer's moat is not the workout. Anyone can generate a workout. ChatGPT can generate a workout. The moat is the relationship, the documented system that delivers the relationship consistently, and the recurring-revenue infrastructure that monetizes it. The 20 documented systems are the moat, not the programming.
2. AI tools should be used aggressively in the back office. The trainers who win this decade will use AI for client comms drafts, content generation, marketing copy, screening question development, and admin automation — freeing their hours for the in-person delivery that is the actual product. Refusing to use AI is sentimentality, not strategy.
3. The high-touch end of the market becomes more valuable, not less. As cognitive services commoditize via AI, premium pricing flows to the categories AI cannot deliver. Trainers in the $150–$300/session in-home market are positioned exactly where the premium accrues. Trainers in the $50–$80/session gym-floor commodity market are competing against an AI that gives the same programming for free.
4. Online coaching is structurally weaker than in-home coaching. This is counterintuitive because online coaching scales better. But it scales better into a market that AI is rapidly entering. An online coach selling $200/month programming-and-check-ins is competing against an AI app selling the same product for $20/month with better adherence tools. An in-home coach selling $600/month subscription training is competing against nothing that AI can deliver. The premium and the durability both favor the in-home model on a 10-year horizon. I made the in-home structural case in detail elsewhere.

Where to Start

If this analysis lands, the moves are unchanged from the moves I have been describing across this site for two years. They are now urgent rather than aspirational.

First: Stop arguing about whether to leave the gym. The trajectory has decided that question for you on a five-to-ten year horizon. The question is the timeline of your transition, not whether one is needed. The independence playbook covers the readiness criteria and the pre-exit timeline.

Second: Build the in-home model with subscription billing, screened clients, and documented systems. The categorical protection only applies if you actually have an in-home subscription business. A solo trainer selling per-session work to whoever walks in is not in the protected category. A documented operator with retained clients and a recurring revenue stack is. The structural moat I described in the playbook is the load-bearing piece.

Third: Use AI in the back office. Generate first drafts of client communications. Use it for content. Use it for screening questions. Use it for marketing copy. Then use the hours you saved to do more in-person training, more relationship-building, more in-home sessions. The shift in your hours from cognitive work to embodied work is the move that compounds.

Fourth: Stop thinking about your career on the timeline a 1980s career counselor would have used. The next decade does not look like the last one. The trainers who treat 2026 like 1998 will spend the decade being confused that the rules they were taught don't work. The trainers who treat 2026 like the structural inflection point it actually is will be in a different business entirely by 2030.

Frequently Asked Questions

About the Author

Jesse Ray Snyder started at Crunch Fitness in San Francisco making $30/hour while sleeping in a 2003 Toyota Tundra. He became their highest-producing resigner within months, left, and built Monterey Personal Training from zero—hitting $9,200 in monthly revenue within five months with no paid advertising. He later scaled to $13,000/month with a second trainer, then deliberately scaled back to ~6 hours/week because the system gave him the freedom to optimize for lifestyle instead of maximum revenue. Across six years of Stripe subscription billing: zero chargebacks, 25-month average client retention (industry average: 3–5 months), and 35+ five-star reviews with zero below five stars. He holds a B.S. in Exercise & Sport Science from Oregon State University (6 years, 4 transfers), is a NASM Corrective Exercise Specialist, a self-taught real estate investor, and serves as a guest lecturer at California State University, Monterey Bay. He consulted for tech startups that went on to nine-figure annual revenue. He is the creator of The Trainer Blueprint.

Follow Jesse: LinkedIn · Instagram · Threads · X · Substack · Reddit