You cannot stop the proliferation of AI, but you can shape how it proliferates and who it benefits.

The prospect of AGI – artificial general intelligence – makes business and tech people alike giddy. Achieving AGI would be an incredible technical feat and massively reduce labor costs for executives and business owners, ultimately enabling them to increase profits.

However, many people are justifiably concerned, frustrated, and even angered by the rapid advancements in and adoption of artificial intelligence technology. They may refuse to use AI or call for government intervention through regulation. Meanwhile, the largest technology and AI companies are making concerted - and sometimes coordinated - efforts to use AI to reduce the size of their human workforces.

Companies use human workers' outputs to improve the AI systems designed to replace those same humans through a process called RLHF – Reinforcement Learning with Human Feedback. Astonishingly, many workers are literally helping these AI companies engineer their own obsolescence, increasing the likelihood of downward mobility for themselves and their families.

History has proven time and time again that innovations that increase productivity will find ways into the market. The question isn't whether AI will transform work, it's whether that transformation will empower human workers or eliminate them. Who will win? Humans or hardware?

The Business Model Trap

Companies that train foundation models face a fundamental conflict of interest rooted in their business model. Think of it like this: if you owned a factory that made both robots and employed human workers, and every robot you sold could replace ten workers somewhere, what would you focus on improving? The robots, of course - that's where the most money is.

Their incentive is simple: the more tasks their AI can perform without human intervention, the more valuable their models become to business owners looking to cut costs. Every improvement to their AI systems increases their potential to automate away human jobs.

When workers correct AI outputs or provide training data, they're inadvertently teaching these systems to perform their jobs without them. It's like training your own replacement at work, except you're doing it for free and at massive scale. Make no mistake: where billions of dollars in potential profits are involved, these foundation model companies will pursue whatever path maximizes the value of their models - and that path leads directly through reducing costs for business owners with full workforce automation.

The Solution: Per-Seat Business Model

This is why we desperately need companies that operate on fundamentally different business models - companies that don't train foundation models at all but instead focus solely on helping human workers leverage AI. These companies typically operate on per-seat or per-user pricing models. Their revenues - and in turn the businesses themselves - directly depend on the number of human workers using their tools.

This creates a crucial tension: while these companies still depend on foundation models for their underlying capabilities, their business success is tied to maximizing employment, not minimizing it.

Every worker who loses their job to automation is a lost customer and every human made more productive is a retained and valuable user.

It's like the difference between a gym that profits from your membership versus a weight-loss surgery clinic. One needs you to keep coming back. The other profits from a one-time elimination of the problem. Except, in the latter case, you don't benefit from the weight-loss, you are the "weight" being lost.

Creating Protective Layers in the AI Ecosystem

Companies focused on human-AI collaboration serve several critical functions:

1. Data Protection: Just as VPNs protect your browsing data from websites and ISPs, these intermediary companies act as a protective layer between you and foundation model companies. They can aggregate, anonymize, and control how worker data flows back to model trainers and, at scale, they have more leverage and incentive to impose stricter privacy and data control requirements on upstream foundation model providers. When you use ChatGPT directly, OpenAI sees everything and, in many cases, you give them an explicit license to use your inputs to improve their models. When you use it and other models through a worker-focused platform, your individual patterns get mixed with thousands of others, making it harder for the foundation model company to make sense of the information.

2. Model Diversification: Android created cheaper and more open alternatives to iPhones, forcing Apple to keep their prices somewhat reasonable and, more importantly, make strong commitments to privacy. Android is largely the reason why Apple continues to innovate by offering more privacy enhancements for customers. Worker-centric AI companies do the same by using multiple foundation models. Like how Android phones can choose between different app stores and services, these platforms can switch between GPT, Claude, Gemini, and others. This prevents any single model provider from gaining monopolistic power and keeps them competing on features that benefit workers.

3. Worker-Centric Features: Unlike foundation model companies that might offer seemingly worker-centric tools and platforms as loss leaders (like free samples at a grocery store meant to get you to buy more), companies that do not train models invest fully in features that maximize human productivity and satisfaction. Human productivity and satisfaction are their only paths to profit. It's the difference between a car manufacturer's basic warranty and a dedicated mechanic shop - one sees service as a cost of doing business and the other sees it as their entire business.

4. Advocacy: With their business success tied to worker employment, these companies become natural advocates for responsible AI development. Like how unions advocate for worker rights because their power depends on membership, worker-centric companies fight for human-centric AI because their revenue depends on human users.

   
   

The Competitive Ecosystem Effect

Supporting worker-centric AI companies creates a ripple effect that benefits everyone. When multiple companies compete for worker subscriptions rather than racing to automate jobs away, it fosters greater fairness and increased competition among foundation model companies. This competition creates a healthier ecosystem focused on increasing worker productivity rather than eliminating workers altogether.

Think of it like the organic food movement. When enough consumers chose organic, it didn't just support organic farmers - it forced conventional agriculture to adopt better practices. Similarly, when workers choose productivity-focused AI tools, it signals to the entire industry that there's money in empowerment, not just automation. This market pressure can redirect even the largest AI companies toward developing features that enhance human capabilities rather than replace them.

Choose Your AI Partner Wisely

Not all per-seat companies are created equal. Some companies offer per-seat pricing options but simultaneously pursue AGI (Artificial General Intelligence) as their stated business goal - a technology that, by definition, would replace human workers entirely.

When choosing AI tools, look for companies that are fully and completely committed to human-AI collaboration as their core mission, not just as a temporary business model.

Would you rather buy ethically-raised meat from a company known for animal welfare violations but offering a "humane" product line, or from a company built entirely around ethical treatment? The same principle applies to AI tools.

Choose companies whose entire existence depends on keeping humans productively employed, not those hedging their bets by playing both sides.

This isn't about what's best for business owners - they'll naturally gravitate toward whatever reduces costs, just as they've always chosen cheaper production methods throughout history. This is about workers making strategic choices about which AI tools to adopt and support.

When you use AI tools from foundation model companies directly:

• Your corrections and feedback can directly train their models to replace you (like teaching a security camera system to spot thieves, then finding out it's being used to monitor employee productivity)

• You're supporting a business model that profits from your obsolescence

• You're accelerating the concentration of power in companies incentivized to eliminate jobs

When you use AI tools from per-seat companies focused solely on worker productivity:

• You're supporting businesses that only succeed if you keep your job (like how Microsoft Office's success depends on office workers existing)

• You're adding protective layers between your work and companies with a direct incentive to re-train models

• You're voting with your dollar for a future where AI empowers rather than replaces

• You're contributing to a competitive ecosystem that keeps foundation model companies focused on productivity enhancement rather than job elimination

The Path Forward

Even if AGI eventually arrives, workers who choose their AI tools thoughtfully can face that future with a clear conscience, knowing they didn't passively accept their own obsolescence. Just as labor movements of the past fought for the 40-hour work week and workplace safety standards - battles that seemed impossible at the time - today's workers can shape how AI develops by being deliberate about which companies they support.

Every subscription to a worker-centric AI platform is a vote against the potential inevitability of mass unemployment. Just because you think a storm is coming, doesn’t mean that you should give in and help to make the storm come faster and with more force.

Every piece of feedback given to a company that depends on your continued employment is an investment in a future where humans and AI work together. Win or lose, at least we'll know we didn't train our replacements without a fight.

We stand at a crossroads. The technology itself is neutral - a knife can be a surgical tool or a weapon. It's the business model around the technology that determines whether AI becomes a tool for human flourishing or obsolescence. Foundation model companies aren't inherently evil, but their business incentives push them toward automation. Companies with per-seat models aren't inherently virtuous, but their incentives align with keeping humans employed and productive.

As workers, we can:

1. Choose AI tools from companies whose business models depend on our continued employment

2. Understand that every interaction with AI is a vote for a particular future - like choosing between shopping at a local store versus a mega-mart

3. Support companies that create protective layers between us and companies that train foundation models

4. Spread awareness that workers have power in shaping AI's development through tool choice

5. Recognize that our collective choices can create a competitive ecosystem that benefits workers

6. Select companies fully committed to human-AI collaboration, not those hedging with AGI ambitions

The proliferation of AI is inevitable, but whether it leads to mass unemployment or mass empowerment is still being decided. That decision isn't made in corporate boardrooms alone - it's made every time a worker chooses which AI tool to use. In the late '90s and early 2000s, consumers created the open internet. Decades later, they've preserved its openness and accessibility.

While you cannot get rid of AI altogether, you can shape the future of work by carefully and thoughtfully choosing which AI tools you use.

Choose tools from companies that need you to succeed. Choose humans over hardware. Your job, your future, and your family's future may depend on it.