When Intelligence Approaches Zero Cost: Rethinking Productivity, Employment, and Prosperity

The Marginal Cost of Intelligence

Artificial intelligence is often discussed as a tool for efficiency, automation, or competitive advantage. Its deeper significance may be civilizational: AI is rapidly reducing the marginal cost of producing some AI-mediated cognitive outputs, including analysis, design, tutoring, translation, coding, and decision support. This does not make intelligence itself costless. Reliability, contextual understanding, verification, responsibility, and judgment still require people, institutions, data, infrastructure, and time. The economic question is therefore not simply how firms can become more productive, but how society might direct increasingly scalable expertise toward broader prosperity.

This is the premise behind what I call Zero World: a normative framework for evaluating how AI-enabled abundance could be directed toward reducing avoidable suffering, expanding opportunity, and moving humanity closer to “zero” on major deficits—poverty, preventable disease, exclusion from education, unnecessary unemployment, hunger, and barriers to human potential.¹² Zero World is not a prediction that cheaper AI will automatically produce these outcomes. It is a lens for asking who benefits, who bears the risks, and what policies and institutions are needed to convert technical abundance into human capability.

The Marginal Cost of Intelligence

In classical economics, land, labor, capital, and physical resources are scarce. Expert knowledge has also been scarce, limited by geography, income, language, and institutional affiliation. A student without a tutor, a small entrepreneur without a consultant, or a local government without data scientists has faced real economic constraints.

The first creation of advanced AI remains expensive, requiring data, chips, energy, research talent, and infrastructure. Once deployed, however, the cost of generating one more explanation, prototype, translation, or learning pathway can be low. AI therefore makes some forms of expertise dramatically more scalable: costly to create, but relatively cheap to replicate. This extends the non-rival character of ideas described by Romer³ and the falling cost of prediction discussed by Agrawal, Gans, and Goldfarb.⁴

Yet low-cost output is not the same as low-cost capability. Reliable economic value may require proprietary data, integration with existing systems, workflow redesign, cybersecurity, compliance, capital, skilled users, and human review. Nordhaus⁵ similarly cautions that accelerating digital capability does not automatically yield equally distributed prosperity. Hidden costs include energy use, data extraction, labor displacement, bias, misinformation, and concentration of power.

Productivity Beyond Automation

The dominant AI-and-work narrative asks which jobs will disappear. Jobs, however, are bundles of tasks involving judgment, communication, coordination, care, creativity, and accountability. AI may automate some tasks while increasing the value of others. Autor⁶ and Acemoglu and Restrepo⁷ show why technological change can both displace labor and create new tasks.

Recent evidence supports a more conditional interpretation. Brynjolfsson, Li, and Raymond⁸ found substantial gains in customer-support productivity, especially among less experienced workers. A 2025 OECD review concludes that effects vary by task, user experience, expertise, and the quality of human-AI collaboration; gains are strongest for well-defined tasks, while complex work still depends on contextual knowledge and oversight.⁹ The ILO’s 2025 update estimates that one in four workers are in occupations with some exposure to generative AI, but emphasizes that most jobs are more likely to be transformed than eliminated because human input remains necessary.¹⁰ These findings do not establish a universal productivity dividend. They show that outcomes depend on what the system does, who uses it, and how work is reorganized.

A small business owner may use AI for market research; a teacher may create learning materials; a nonprofit may draft a grant; a farmer may access climate information; and a student may receive tutoring in another language. These are plausible pathways to wider participation, not guarantees. Productivity matters partly because it expands what people can do, but only when access, skills, and institutional support accompany the tool.

Economic Liberation and the Zero World Vision

Sen¹¹ defines development as expanding people’s capabilities to live lives they have reason to value. Keynes¹² likewise asked how abundance might change work, leisure, income, and meaning. AI makes these questions immediate. If certain cognitive outputs become cheaper, the benefit should not accrue only as faster production or higher margins; it could become more capability, opportunity, and freedom from avoidable deprivation.

In a Zero World, AI might support underserved students, rural health navigation, affordable legal information, entrepreneurship coaching, or translation for civic participation. But the framework is explicitly normative: these applications require choices about public investment, access, safeguards, and accountability. AI could instead intensify inequality if advanced models, data, cloud infrastructure, and distribution channels remain concentrated. Intelligence could become cheaper to generate but more expensive to access, or simply rented from a small number of providers. Zero World therefore rests on access, literacy, and accountability—not technological inevitability.

Emerging Business Models

Lower-cost cognitive outputs may enable new business models, but these should be treated as emerging possibilities rather than settled economic laws. “Service-as-software” describes systems that deliver outcomes—such as a report, customer resolution, lesson, or compliance review—rather than merely providing a tool. AI may also lower entry barriers for micro-enterprises by helping small teams perform research, design, coding, finance, marketing, and operations. Whether these firms survive will still depend on distribution, trust, capital, domain knowledge, and the ability to deliver reliable outcomes.

Purpose-driven AI ventures could organize around deficits such as educational gaps, healthcare delays, food waste, financial exclusion, or workforce mismatches. Trust markets may also expand as AI-generated content and decisions multiply: verification, privacy protection, auditing, certification, and explanation could become valuable services. These are hypotheses about where demand may grow, not inevitable shifts. In each case, the scarce resource may move from producing an answer to validating it and taking responsibility for its consequences.

Employment and the Human Premium

AI is likely to change employment unevenly. The value of human work may increasingly lie in judgment, empathy, ethics, cultural understanding, leadership, creativity, and responsibility—the ability to decide what matters, whose interests are affected, and what should be done. This “human premium” is not a claim that people possess an immutable advantage; it is a reminder that accountable social and organizational decisions cannot be reduced to inexpensive output.

Education should therefore teach students to prompt, evaluate, verify, challenge, and govern AI, alongside collaboration, moral reasoning, communication, resilience, and imagination. The future of work may favor people who collaborate effectively with AI while retaining the capacity to question it.

Policy Implications

A Zero World policy agenda can be condensed into four connected priorities. First, treat safe AI access as part of digital public infrastructure for schools, libraries, nonprofits, small businesses, and local governments. Second, make AI literacy a civic capability, including awareness of failure, bias, misinformation, privacy, and overreliance. Third, preserve competition so that models, data, compute, and distribution do not become intelligence monopolies. Fourth, adapt labor and high-impact-domain governance through lifelong learning, transition support, worker voice, transparency, auditability, human oversight, and clear accountability.

Conclusion

Zero economics does not claim that everything becomes free. It recognizes that when some AI-mediated cognitive outputs become dramatically cheaper, society has an opportunity to redesign the relationship between technology and prosperity. The outcome is open. Lower costs could support broader liberation, or they could strengthen concentration and exclusion. Zero World is a normative test: does deployment reduce vulnerability, expand capability, and distribute meaningful gains?

When intelligence approaches zero cost, wisdom becomes priceless.

References

¹ Gopalkrishnan, S. S. (2025). The Zero World: Envisioning a future where AI eliminates global challenges. In S. S. Gopalkrishnan & J. J. Gonzalez III (Eds.), The world remade by artificial intelligence. McFarland.

² TEDx Talks. (2025). How AI could address our greatest challenges | Shalini Gopalkrishnan | TEDxCSTU [Video]. YouTube.

³ Romer, P. M. (1990). “Endogenous technological change.” Journal of Political Economy, 98(5), S71–S102. https://doi.org/10.1086/261725

⁴ Agrawal, A., Gans, J., & Goldfarb, A. (2018). Prediction machines: The simple economics of artificial intelligence. Harvard Business Review Press.

⁵ Nordhaus, W. D. (2021). “Are we approaching an economic singularity? Information technology and the future of economic growth.” American Economic Journal: Macroeconomics, 13(1), 299–332. https://doi.org/10.1257/mac.20170105

⁶ Autor, D. H. (2015). “Why are there still so many jobs? The history and future of workplace automation.” Journal of Economic Perspectives, 29(3), 3–30. https://doi.org/10.1257/jep.29.3.3

⁷ Acemoglu, D., & Restrepo, P. (2019). “Automation and new tasks: How technology displaces and reinstates labor.” Journal of Economic Perspectives, 33(2), 3–30. https://doi.org/10.1257/jep.33.2.3

⁸ Brynjolfsson, E., Li, D., & Raymond, L. R. (2025). “Generative AI at work.” The Quarterly Journal of Economics, 140(2), 889–942. https://doi.org/10.1093/qje/qjae044

⁹ Calvino, F., Reijerink, J., & Samek, L. (2025). The effects of generative AI on productivity, innovation and entrepreneurship. OECD Artificial Intelligence Papers, No. 39. https://doi.org/10.1787/b21df222-en

¹⁰ Gmyrek, P., Berg, J., Kamiński, K., et al. (2025). Generative AI and jobs: A 2025 update. International Labour Organization. https://www.ilo.org/publications/generative-ai-and-jobs-2025-update

¹¹ Sen, A. (1999). Development as freedom. Oxford University Press.

¹² Keynes, J. M. (1930). “Economic possibilities for our grandchildren.” In Essays in persuasion. Macmillan.