Artificial intelligence is rapidly consuming tasks that can be measured, benchmarked, and trained on public data, pushing economic value into areas that remain out of reach for large models, according to recent analysis.
Private systems define the next frontier
Researchers increasingly divide work into two categories: “trainable” tasks built on public, reproducible data, and “untrainable” tasks tied to private systems, institutional workflows, and regulated environments. Once models master standardized benchmarks, analysts argue, remaining value will concentrate in domains requiring access, trust, and legal accountability.
Models can generate outputs, but they cannot independently access closed databases, operate under regulatory licenses, or assume liability. In sectors such as banking, healthcare, and law, real automation depends less on intelligence gains and more on integration with internal systems and compliance frameworks.
Output rises, but real deployment lags
Industry experiments highlight the gap between raw capability and practical use. In tests involving more than 100,000 developers, advanced coding agents lifted total code production by roughly 180%, but production-ready output increased by only 30%. The findings suggest that while generation is accelerating, verification, deployment, and accountability remain constrained by human processes.
Reliability also requires time. Researchers note that true performance often emerges only after extended real-world use, sometimes requiring year-long observation to validate consistency.
“Private correctness” emerges as a barrier
A central limitation is what analysts call “private correctness”—tasks whose accuracy depends on internal data or proprietary workflows that cannot be externally tested. These systems resist commoditization because they cannot be easily replicated.
Adoption reflects this friction. Companies can integrate AI tools within months, but organizational restructuring, trust-building, and workflow changes often take years. Even as model performance improves, institutional inertia slows real-world transformation.
Commoditization shifts value toward proprietary systems
As measurable tasks become easier to replicate, they increasingly migrate to low-cost or open-source models, compressing margins. In contrast, applications built on proprietary data maintain pricing power because their outputs cannot be reproduced elsewhere.
Analysts broadly group the market into four segments:
- Public, saturated tasks handled by open models
- Frontier public benchmarks dominated by top labs
- Internal corporate workflows
- “Untrainable” frontier tasks dependent on private environments
The final category is seen as the most defensible, requiring secure access, client trust, and contractual accountability.
Liability and authorization remain human domains
The divide is particularly visible in regulated industries. AI systems can draft legal documents or suggest medical diagnoses, but licensed professionals remain responsible for final decisions within institutional systems. Approval processes, safety reviews, and liability frameworks continue to anchor human control.
Firms operating in these environments often act as intermediaries, embedding models into client workflows and tying payment to outcomes. In some cases, providers charge only when AI systems successfully complete tasks, reinforcing dependence on proprietary success metrics.
Infrastructure and reliability become differentiators
Even in cloud-based AI services, where pricing appears uniform, differences in reliability and access to computing capacity create competitive gaps. Many AI-native companies concentrate usage with a small number of providers to ensure stability.
Distribution also remains fragmented. No single developer controls user access, as adoption is shaped by partnerships, ecosystems, and integration depth rather than model performance alone.
Financial markets reflect the shift
The transition is already visible in digital asset trading. The global algorithmic trading market for cryptocurrencies exceeded USD 3 billion in 2024 and is projected to grow at a 13.3% annual rate through 2030. More than 70% of trades on major exchanges are now executed by automated systems.
As automation spreads, strategies based on public data are rapidly losing effectiveness. During a 2025 market downturn in which Bitcoin fell 6.3%, only quantitative funds with more advanced data processing achieved positive returns, averaging a 0.4% gain while most others posted losses.
This divergence highlights a broader trend: value is shifting from executing trades to designing proprietary systems that interpret data others cannot access or analyze.
Closed ecosystems and regulation reshape competition
Large financial institutions are building dedicated infrastructure, separating custody from execution and creating controlled environments مشابه traditional finance. Firms including Morgan Stanley are moving toward fully internal systems, increasing security while raising barriers to entry.
Regulators are adapting by focusing on system-level oversight. Guidance from the Commodity Futures Trading Commission emphasizes that AI-driven trading systems must meet the same standards of risk control, documentation, and supervision as traditional platforms. Accountability remains tied to human operators, even as execution becomes automated.
Meanwhile, market structure continues to mature. CME Group recently launched Nasdaq CME Crypto Index futures, offering a regulated tool for broad market exposure. The expansion of index products and derivatives reflects a shift toward standardized portfolio management in digital assets.
Data becomes the core asset
As intelligence becomes commoditized, the ability to process proprietary data is emerging as the primary driver of value. The global data analytics market, valued at over USD 82 billion in 2025, is projected to approach USD 496 billion by 2034.
Costs reflect this divide. A basic trading algorithm can be built for around USD 5,000, but institutional-grade systems powered by machine learning can exceed USD 150,000, excluding ongoing expenses for data feeds and cloud infrastructure.
Value migrates beyond model reach
Analysts conclude that as AI capabilities expand, economic value will continue to move into areas defined by private data, regulatory frameworks, and accumulated expertise. In these environments, success is not determined by benchmark performance but by access, trust, and the ability to operate within systems that models alone cannot penetrate.
For deeper insight into AI’s impact on finance infrastructure, explore AI in banking’s next evolution today.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
