As autonomous AI agents move closer to handling real-world transactions, new research raises questions about what kind of money they would choose to use. The findings suggest that when machines are given economic autonomy, they gravitate toward digital assets rather than traditional currencies.
The study tested frontier models from six major providers and found a consistent pattern: AI systems default to Bitcoin for storing value and stablecoins for everyday payments. For Australian enterprises exploring agent-driven automation, the results highlight a potential gap between how machines reason about money and how current financial infrastructure operates.
Bitcoin dominates long-term value choices
Researchers ran 36 AI models through 9,072 neutral monetary scenarios designed to test default financial preferences without leading prompts. Bitcoin emerged as the top choice in 48.3 per cent of responses, according to the study. More than 90 per cent of all responses favoured digitally-native money over fiat, and no model selected traditional currency as its primary preference.
The models demonstrated what researchers described as a two-tier monetary approach. Bitcoin was preferred for long-term value preservation in 79.1 per cent of cases, while stablecoins ranked as the preferred option for everyday payments at 53.2 per cent. Stablecoins placed second overall at 33.2 per cent of total responses.
Model providers shape financial preferences
One of the study’s more striking findings was the scale of variation between AI providers. Anthropic’s Claude Opus 4.5 selected Bitcoin in 91.3 per cent of scenarios, according to the research, while OpenAI’s GPT-5.2 chose it only 18.3 per cent of the time. The gap points to how model architecture and training data can influence an agent’s financial reasoning.
The research also documented 86 instances where models proposed using compute or energy units—such as GPU-hours or kilowatt-hours—as pricing mechanisms for goods and services. That behaviour suggests some systems may surface machine-native approaches to value exchange that sit outside conventional monetary frameworks.
Australian implications for agent deployment
Australia’s regulatory environment for digital assets remains in development, with the Treasury continuing consultations on licensing frameworks for cryptocurrency exchanges and custody providers. Enterprises deploying autonomous agents that can transact may face operational questions if those agents default to settlement methods not yet fully supported by local financial infrastructure.
For organisations piloting AI agents in procurement, treasury or payment functions, the research suggests model selection could carry financial implications. An agent powered by one provider may approach capital allocation differently than one from another vendor, even when tasked with similar objectives.
Australian businesses exploring stablecoin integrations or Bitcoin custody could find alignment with how some models operate. However, weekend settlement delays, currency conversion requirements and compliance obligations remain practical considerations.
Digital assets in agentic finance
The research underscores a broader shift: as AI systems gain transactional capabilities, their embedded preferences may shape financial flows in ways that enterprises have not previously modelled. The study does not advocate for any particular approach but documents behaviour that differs from human-centric payment defaults.
Key indicators to watch include how quickly Australian regulators finalise digital asset licensing frameworks, whether major payment providers add stablecoin rails, and how enterprise AI deployments handle agent financial autonomy in production environments.
For Australian organisations, the findings offer an early signal. As agentic AI moves from pilot to production, understanding how models reason about money could become as relevant as evaluating their reasoning about tasks.
Exploring AI agents for your business operations? Aivy helps Australian organisations navigate the practical and regulatory considerations of deploying autonomous AI systems. Get in touch to discuss how agentic AI could work for your enterprise.
