A flyer handed out at an anti-AI march in London last February channeled the underpants gnomes from South Park: 'Step 1: Grow a digital super mind. Step 2: ? Step 3: ?' The joke cuts deep because it captures the central problem in artificial intelligence today. Companies have built the technology. They promise transformation and profit. But the middle step — how to actually deploy AI profitably — remains a black box.
The Big Picture
The underpants gnomes meme, which first aired in 1998, has become a staple of internet satire. Elon Musk once used it to explain his Mars funding plan. Now it perfectly describes the AI industry's predicament. A study from Mercor, an AI hiring startup, tested agents powered by top models from OpenAI, Anthropic, and Google DeepMind on 480 workplace tasks frequently performed by bankers, consultants, and lawyers. The results were sobering: every agent failed to complete most of its duties. Only 30% of tasks were completed successfully. The gap between lab demos and real-world performance is vast.
“The missing Step 2 in AI — profitable deployment — remains a black hole that neither companies nor regulators have been able to illuminate.”
The lack of a clear Step 2 is not just a technical problem but a governance one. Companies are pouring billions into AI infrastructure without standardized metrics to measure return. Investors face a paradox: AI company valuations soar while studies show agents failing at everyday tasks. Who wins? Consultants selling digital transformation. Who loses? Retail investors buying the hype without evidence of real adoption. Regulators are caught between pressures to innovate and demands for job protection.
By the Numbers
- AI agent failure rate: 70% of AI agents tested by Mercor failed to complete basic banking, consulting, and legal tasks. This includes tasks such as analyzing financial statements, drafting simple contracts, and summarizing legal cases.
- Job impact by sector: Anthropic predicts managers, architects, and media workers will be most affected by LLMs; groundskeepers, construction workers, and hospitality staff, least. Automation is not uniform: repetitive cognitive jobs are in the crosshairs, while roles requiring physical dexterity or complex human interaction are more protected.
- Coding speed vs. strategic judgment: Advances in coding do not translate to better strategic decisions, where LLMs remain weak. A Stanford University study showed that AI models improve in programming tasks at a rate of 40% annually, but in strategic reasoning only 5%.
- Regulatory vacuum: Groups like Pause AI demand regulatory pauses, but there is no consensus on what rules to apply or who would enforce them. The European Union advances with the AI Act, but its effective implementation is scheduled for 2027, leaving a regulatory gap of at least one year.
- Investment without return: Big tech companies including Microsoft, Google, and Amazon have committed over $150 billion combined to AI infrastructure in 2025-2026, according to Goldman Sachs estimates. Yet direct revenue from AI products generates less than 5% of their total revenue.
Why It Matters
The lack of a clear Step 2 creates an information vacuum filled by the wildest claims. Every week, a new viral post announces that AI will replace humans, but concrete data is scarce. Investors face a paradox: AI company valuations soar while studies show agents failing at everyday tasks. Who wins? Consultants selling digital transformation. Who loses? Retail investors buying the hype without evidence of real adoption. Regulators are caught between pressures to innovate and demands for job protection.
Geography also matters. While the European Union advances with the AI Act, the United States lacks a federal framework, and China pushes AI with state control. This regulatory fragmentation adds another layer of uncertainty for global companies. For example, a company deploying AI in its EU operations must comply with transparency and risk assessment requirements, while in the US there are no clear guidelines, creating uneven compliance costs.
What This Means For You
Whether you are an investor, business operator, or professional, the message is clear: hype is not strategy. Real AI implementation requires rethinking workflows, investing in training, and accepting that benefits will take years.
- 1Investors: Demand metrics for real deployment, not just model development. Prioritize companies that show use cases with measurable results, such as cost reduction or productivity gains in specific areas. Be wary of those that only report model capability advances.
- 2Professionals: Invest in skills complementary to AI, such as strategic judgment, hybrid team management, and data ethics. The safest jobs are those requiring human contact or complex decision-making. Consider certifications in AI governance.
- 3Companies: Don't implement AI for fashion. Evaluate where it truly adds value and be prepared to redesign processes, which takes time and resources. Start with pilot projects in low-risk areas, such as internal report automation, before scaling.
What To Watch Next
The coming months will be critical. In June 2026, the European Commission will publish the first detailed guidelines for high-risk AI implementation, potentially setting a global precedent. In the US, the SEC may require companies to report AI adoption metrics in quarterly filings, following the model of sustainability reports. Additionally, big tech Q3 results will show whether AI infrastructure investment is generating returns. Also watch for upcoming funding rounds for applied AI startups: if investors become more cautious, it could signal that the hype cycle is moderating.
The Bottom Line
AI is not a fraud, but it is not the immediate panacea some claim. Step 2 — profitable deployment — remains uncharted territory. Investors and companies that focus on real metrics, regulation, and process adaptation will be better positioned when the dust settles. Until then, the underpants gnomes meme remains the best description of the industry's current state. The question is not whether AI will transform the economy, but when and how the gap between promise and reality will close.
