Today, AI surpasses human intelligence across a wide range of well-defined, isolated tasks, but not as a whole. To reach AGI, Artificial General Intelligence, two main elements are still missing: the ability to gather data that AI cannot access through data centres, and an increased ability to integrate all data. These are necessary to develop the key dimensions of causal understanding, robustness and intentionality. This is the second of three posts on the development of AGI and some of the risks it entails for the economy, governance, democracy and international cooperation.
The blog series is structured as follows:
- What is missing to reach AGI? (previous post)
- Impact on the economy and governance (this post)
- The democratic and political risks of AGI (next post)
AGI challenges two fundamental constraints of modern economies
At its core, today’s economy and financial system rest on two key assumptions that are particularly relevant for the development of AGI:
- Trust as a foundation. Mutual trust between actors is a prerequisite for fiat economies and for credit. AGI will surpass human intelligence in both breadth and flexibility, cf. the previous post. As a result, AGI will be highly persuasive and may therefore challenge trust between individuals, while also making it more difficult to distinguish between genuine and synthetic decision-making foundations.
- Intelligence as a scarce resource. Economic value creation today is largely constrained by the scarcity of skilled labour and decision-making capacity. This scarcity disappears as development passes AGI and continues towards ASI (Artificial Superintelligence).
AGI may therefore lead to significantly higher productivity and reduced need for human labour. It also reshapes the very notion of capital, as access to compute, data, and models increasingly becomes the decisive means of production. Taken together, this may lead to new forms of growth and distribution, potentially rendering existing economic models and policies insufficient.
AGI may create a structural shift in productivity, …
The key point is the direction rather than the magnitude of current gains. With autonomous systems capable of executing entire workflows, we are structurally moving from improving tools to substituting cognitive functions.
Already today, AI can deliver substantial productivity gains in knowledge work, particularly where tasks are well defined. AI has, for example, demonstrated the ability to generate hypotheses, design and conduct experiments, evaluate results, and iterate without human intervention.
This fundamentally changes the production function. Labour becomes more efficient and, in some cases, redundant. Productivity growth is therefore no longer necessarily tied to education, experience, or organisational complexity. Going forward, growth will increasingly depend on access to compute, data, and effective AI systems. Ultimately, it will also depend on access to energy and physical infrastructure, which define the real limits to scaling.
… as a feedback loop from the automation of innovation may emerge
As AI enhances research and development, a feedback loop emerges in which AI improves research, which in turn improves AI. This may lead to very high innovation velocity, shorter technological cycles, and increased competitive intensity.
As a result, the bottleneck for innovation may shift from the production of knowledge to the validation of knowledge, cf. for instance Anthropic. This increases the importance of institutions and mechanisms that can ensure verification.
Causal understanding is central to AGI, as discussed above, but normative judgement will still require human involvement for many years to come.
Governance may constrain certain outcomes, …
The future is therefore shaped by multiple possible structures rather than a single, deterministic path. The World Economic Forum (WEF), for example, outlines four broad trajectories towards 2030:
- High productivity, high inclusion, where AI drives broad-based prosperity and gains are relatively evenly distributed
- High productivity but low inclusion, where growth is strong but concentrated among a few actors with access to technology and data
- Fragmented development, where geopolitics and regulation limit access to technology, reducing overall gains
- Low trust and low growth, where lack of trust in systems and institutions constrains the use of AI and thus economic growth
Across all scenarios, the technological potential is substantial, but outcomes are constrained by governance. At the same time, the speed of development is critical: the faster technology advances relative to institutional adaptation, the greater the risk of economic and political imbalances.
… but may need to be rethought
AGI thus challenges not only the economy but also the very structure of governance. Three shifts are particularly important:
- From regulation to architecture. When regulation proves insufficient, systems must be designed with embedded control, transparency, and verifiability.
- From national to transnational frameworks. When AI systems operate globally while regulation remains national, a structural mismatch emerges that becomes harder to manage as complexity increases.
- From trust to verification. When AI can convincingly simulate expertise, subjective trust becomes insufficient and objective verification becomes necessary.
In addition, decision-making processes gradually shift as AI becomes part of the decision basis. This may quickly evolve into de facto automated decision-making, where human oversight is formally maintained but effectively eroded. This is already visible in tensions between public authorities and AI developers, for example regarding the use of AI in autonomous systems.
The key question will therefore likely be whether decisions are perceived as legitimate rather than whether they are correct.
The distributional impact may be the most underestimated consequence
If intelligence suddenly becomes an abundant resource, distribution mechanisms change fundamentally. This raises several questions:
- Who owns the means of production (data, models, compute)?
- How is value creation distributed?
- What is the role of humans in the value chain?
These means of production exhibit strong scale and network effects, which may lead to a high concentration of economic power. This concentration may be difficult to reverse, given its breadth and depth.
Historically, economies have absorbed technological shifts through the creation of new job types and sectors. With AGI, however, it is not given that new tasks will emerge at the same pace as old ones disappear. The speed of adjustment in labour markets and institutions may also be significantly slower than the pace of technological development. Historically, such imbalances have often manifested as financial crises, political fragmentation, or institutional erosion.
In sum, AGI has deep structural implications for the economy
AGI thus has structural implications for the fundamental assumptions of the economy: trust between actors and the scarcity of intelligence. The central question is whether institutions, governance, and distribution mechanisms can keep pace. In practice, development is likely to be driven by a limited number of actors with access to capital, compute, and data, reflecting a corresponding concentration of decision-making power.
The economy may adapt to the technology. Governance and legitimacy are far more uncertain.
The next blog post outlines some of the broader implications of AGI for democracy and political stability.