Imagine you’re a top-performing call center agent. Your ability to calm frustrated customers and solve complex problems has made you invaluable to your company. Now imagine that same company uses recordings of your best conversations to train an AI assistant that helps junior colleagues perform at your level. You’ve just become training data – and that transformation carries real risks for your career and compensation.
Recent research from MIT reveals a troubling paradox in AI adoption. While studies show AI assistants significantly improve productivity – helping call center agents resolve problems more effectively and junior developers complete tasks faster – these gains come at a cost to the workers whose expertise fuels the systems. The AI models that promise to democratize skills are essentially digitizing human expertise, allowing companies to scale individual capabilities across their organizations.
The Expertise Extraction Economy
This dynamic extends far beyond call centers. Across industries, daily work now produces rich digital traces that serve as training data for AI systems designed to perform the same tasks. Consulting firms fine-tune models on past engagements, software teams train assistants on internal code, and sales organizations mine call logs. Even creative industries draw on archives of past work to guide generative tools.
“Historically, economic security has rested on the scarcity of skill,” notes the MIT research. “People invest heavily in education and on-the-job learning precisely to acquire capabilities that are rare and difficult to replicate. That rarity underwrites higher wages and bargaining power. Generative AI alters this logic.”
The Copyright Conundrum
This expertise extraction isn’t limited to internal company data. The creative industries are fighting their own battle against AI training practices. As reported by the Financial Times, tech companies are facing increasing legal challenges over using copyrighted material without permission to train AI models. The New York Times is suing Microsoft and OpenAI for using its journalism to train ChatGPT, while a German court ruled it illegal to use copyrighted song lyrics to train generative AI models without a license.
Anthropic paid $1.5 billion to settle a class-action lawsuit by book authors over unauthorized training data use, and Amazon won an injunction against Perplexity AI for allegedly illegally scraping its website. These cases highlight a fundamental tension: AI systems require immense quantities of human-created content for training, but creators are increasingly demanding compensation and control over how their work is used.
Identity Theft in the Digital Age
The problem extends beyond copyright to personal identity. TechCrunch reports that journalist Julia Angwin has filed a class action lawsuit against Grammarly’s parent company Superhuman for using her and other experts’ identities without consent in an AI feature called ‘Expert Review.’ The feature simulated editorial feedback from personalities like Stephen King and Kara Swisher, available to subscribers paying $144 per year.
“I have worked for decades honing my skills as a writer and editor, and I am distressed to discover that a tech company is selling an imposter version of my hard-earned expertise,” Angwin stated. Tech journalist Kara Swisher responded more bluntly: “You rapacious information and identity thieves better get ready for me to go full McConaughey on you. Also, you suck.”
The Productivity Paradox
Despite these challenges, AI continues to drive productivity gains across industries. IndustryWeek analysis reveals that while AI faces a ‘productivity paradox’ similar to the PC revolution – with fewer than half of AI projects moving beyond pilot phases despite substantial investments – the biggest gains come from cross-functional AI integration.
AMD’s generative AI system cuts time for a 14-step process by about 90%, saving more than 3,100 staff hours annually. “Productivity only soared when PCs became interconnected across organizations,” the analysis notes, suggesting that similar integration will unlock AI’s full potential.
Strategic Alliances and Market Realities
The AI landscape is also seeing strategic shifts that could reshape how expertise is valued. The Financial Times reports on a strategic alliance between Microsoft and Anthropic, where Anthropic’s general-purpose AI agent Cowork will be integrated into Microsoft’s AI assistant Copilot. This partnership comes as Microsoft’s Copilot has had underwhelming adoption with only 15 million paid seats (3% of Office users), while Cowork has gained traction as a leading example of AI agents for work tasks.
For workers, this integration means their expertise could be leveraged across even broader platforms, raising questions about compensation and control. The integration allows Cowork to operate in the cloud with full access to Microsoft’s data and governance frameworks, potentially making AI assistance more useful – but also more pervasive.
Navigating the New Labor Market
So what should professionals do in this new reality? The MIT research suggests three strategic shifts:
- Rethink productivity: Workers need to consider how much of their job process they share with employers. If AI training is likely to weaken their position, sharing less may be sensible; if it strengthens their role or pay, sharing more may be worthwhile.
- Rethink competition: Data makes the market for expertise global. Anyone who can generate similar outputs becomes someone who can generate the data necessary to build an AI model capable of doing your work.
- Rethink cooperation: “Workers can end up competing against one another by supplying data too cheaply to companies or intermediaries that recruit people to train AI to do their jobs,” the research warns. Collective action may be necessary to ensure workers share in AI’s gains.
The challenge for businesses is equally complex. As IndustryWeek notes, manufacturers must build digital foundations with three steps: robust data infrastructure, AI-ready applications, and adopting evolving AI models. But they must also navigate the ethical and legal implications of using worker expertise as training data.
As AI systems become more sophisticated, the line between tool and competitor blurs. The workers whose expertise trains these systems face a fundamental question: Are they building tools that enhance their value, or creating digital replacements that could undermine their careers? The answer may determine not just individual career trajectories, but the future of work itself.