Finance
Artificial intelligence and asset pricing: The power of transformers
The rapid advancement of artificial intelligence (AI) has reshaped numerous fields, and finance is no exception. Asset pricing, a domain traditionally dominated by linear models and factor-based approaches, is now experiencing a transformative shift due to AI’s capacity to uncover persistent predictive patterns in financial data. The introduction of large-scale AI models – particularly transformer-based architectures – has significantly enhanced our capability to model complex relationships among assets and firms (Eisfeldt et al. 2023), leading to improved forecasting and risk assessment.
In this column, we explore how AI-driven asset pricing models, particularly those incorporating transformers, leverage cross-asset information sharing to reduce pricing errors and enhance predictive accuracy. These innovations offer a novel perspective on how financial markets process information and determine asset prices.
Traditional versus AI-driven asset pricing
Most traditional asset pricing models, such as the Fama and French (1993) framework, rely on predefined factors that explain asset returns. While effective, these models assume a fixed and linear relationship between asset characteristics and expected returns. AI, on the other hand, introduces a non-linear, data-driven approach, identifying patterns that are often invisible to traditional methods.
Machine learning models, including tree-based methods and neural networks (Gu et al. 2020), have improved asset return prediction by capturing complex relationships between firm characteristics and returns. However, these models typically focus on ‘own-asset prediction’, meaning that they use only an asset’s individual characteristics to estimate future returns. This approach ignores the broader context in which assets interact.
The role of transformers in asset pricing
In a recent paper (Kelly et al. 2024), we introduce the Artificial Intelligence Pricing Model (AIPM), which embeds transformer networks into the stochastic discount factor (SDF) framework. While initially developed for natural language processing, we show that transformers are remarkably effective in financial applications due to their ability to capture cross-sectional dependencies across assets.
Unlike traditional machine learning models, transformers incorporate the ‘attention mechanism’, allowing them to dynamically adjust the weight placed on different inputs based on their relevance. In the context of asset pricing, this means that the model not only considers an asset’s own characteristics but also how these characteristics interact with those of other assets. This approach significantly enhances predictive power by leveraging market-wide information.
Empirical findings: Performance of AI-based models
We evaluated the performance of the transformer-based AIPM using a dataset of US stock returns and conditioning variables. Compared to traditional asset pricing models and other machine learning approaches, the AIPM demonstrated:
- Lower pricing errors: The model achieved significantly smaller out-of-sample pricing errors compared to traditional factor-based models and neural networks without attention mechanisms.
- Higher Sharpe ratios: By integrating cross-asset dependencies, the transformer-based model outperformed existing approaches in terms of risk-adjusted returns out-of-sample.
- Scalability and complexity gains: We found that increasing model complexity – by incorporating deeper transformer layers – consistently improved predictive performance. This supports the notion that AI models benefit from higher parameterisation when applied to asset pricing.
Empirical evidence from the literature on large language models (e.g. Kaplan et al. 2020) indicates that adding more transformer blocks enhances the capacity of models to effectively represent language. Models with deeper architectures are able to capture more abstract features and longer-range dependencies than shallower models. Each additional layer refines the attention distributions, allowing the model to consider both short-term and long-term relationships. Interestingly, the same benefits of transformer complexity emerge in the context of the asset pricing model.
These findings suggest that AI-driven asset pricing models are more efficient in processing vast amounts of financial data, leading to more accurate and robust predictions.
Implications for investors and policymakers
The integration of AI into asset pricing has profound implications for market participants and policymakers. Investors can benefit from improved portfolio allocation strategies driven by AI’s capability to identify subtle pricing inefficiencies, a concept aligned with research on AI and personality traits shaping economic returns (Makridis 2025). Meanwhile, regulators and policymakers must consider the impact of AI-driven trading on market stability and efficiency.
Moreover, the adoption of AI in asset pricing challenges traditional views on market efficiency, echoing broader concerns about AI’s macroeconomic impact and productivity gains (Filippucci et al. 2024). If AI models consistently outperform classical frameworks, it may suggest that markets are less efficient than previously thought, opening the door for further research into the nature of pricing anomalies.
Conclusion
The fusion of artificial intelligence and finance is revolutionising asset pricing. Our research demonstrates that transformer-based models significantly enhance return predictions by leveraging cross-asset information sharing. As AI continues to evolve, its role in financial decision-making will only grow, offering new opportunities for investors and reshaping our understanding of market dynamics.
References
Gu, S, B Kelly and D Xiu (2020), “Empirical Asset Pricing via Machine Learning”, Review of Financial Studies 33 (5): 2223-2273.
Eisfeldt, A, G Schubert and M B Zhang (2023), “Generative AI and Firm Valuation”, VoxEU.org, 4 June.
Fama, E and K French (1993), “Common risk factors in the returns on stocks and bonds”, Journal of Financial Economics 33(1): 3-56.
Filippucci, F, P Gal, and M Schief (2024), “Miracle or Myth? Assessing the Macroeconomic Productivity Gains from Artificial Intelligence”, VoxEU.org, 8 December.
Kaplan, J, S McCandlish, T Henighan et al. (2020), “Scaling laws for neural language models”, arXiv preprint arXiv:2001.08361.
Kelly, B Kuznetsov, S Malamud and T Xu (2024), “Artificial Intelligence Asset Pricing Models”, NBER Working Paper No. w33351
Makridis, C (2025), “The Role of Personality Traits in Shaping Economic Returns Amid Technological Change”, VoxEU.org, 31 January.
Vaswani, A, N Shazeer, N Parmar, J Uszkoreit, L Jones, A N Gomez, L Kaiser and I Polosukhin (2017), “Attention is all you need”, Advances in Neural Information Processing Systems 30.
For a company navigating one of the most consequential transformations in its history, financial clarity is not optional—it is essential. Natura &Co, the Brazilian personal care and cosmetics group behind iconic brands such as Natura and Avon, has long been committed to combining purpose-driven business with commercial performance. After a period of strategic portfolio reshaping, including the divestiture of its Aesop and The Body Shop holdings, the company is now sharpening its focus on profitability and operational excellence across Latin America and global markets.
At the center of that effort sits a deceptively complex challenge: understanding, in real time, which revenue and cost factors are driving or eroding gross margin across a highly diversified business. For years, answering that question meant manual reporting, delayed insights, and finance teams spending valuable time on data gathering rather than analysis.
That’s now changing, thanks to a co-innovation initiative developed together with SAP and Numen, a global SAP partner specializing in digital transformation and enterprise software implementation.
From manual reporting to proactive decision intelligence
The project’s goal was to replace a labor-intensive gross margin analysis process with a generative AI application embedded directly into Natura &Co’s financial workflows. Built on SAP Business AI Platform, SAP’s unified foundation integrating business technology, data, and AI capabilities, the application connects directly to data in SAP S/4HANA to provide finance teams with automated insights and narrative recommendations in real time, without the need for manual data pulls or offline reporting.
The application enables users to explore revenue, cost, and margin drivers interactively, identifying at a glance which elements are protecting or eroding margin performance across markets and product lines. Crucially, human oversight remains central to the design: the AI application generates insights, while finance professionals retain full control over interpretation and decisions.
“The implementation of gross margin analysis using AI in SAP S/4HANA marked an inflection point in the analytical capability of our finance area,” said Rogério Dias Garcia, tech manager, ERP Latam, Natura &Co. “We overcame delays and raised the standard of insights by integrating margin analysis from SAP S/4HANA with a large language model connected via the SAP AI Core layer. This architecture allowed us to provide, in an agile, secure, and completely anonymous manner, a stratified and precise view of gross margin offenders and protectors—discriminating exactly which revenue or cost elements were driving market performance.”
A collaborative architecture for scalable AI adoption
Natura &Co’s application derived from a prototype SAP partner Numen created in early 2024 at SAP’s global Hack2Build on business AI, leveraging the generative AI capabilities of SAP Business AI Platform. The solution was designed and developed through close collaboration between Natura &Co, Numen, and SAP. From the outset, the approach was to align AI adoption with concrete business priorities, ensuring the application would be scalable and production-ready rather than a standalone prototype.
Numen brought deep SAP implementation expertise to the project, combining knowledge of SAP S/4HANA architecture with hands-on experience in building solutions on SAP Business AI Platform. The technology stack—SAP S/4HANA, SAP AI Core, SAP Fiori, and SAP Business Technology Platform—provided the secure, integrated foundation needed to connect financial data with generative AI capabilities in an enterprise context.
“SAP enabled the transformation by providing the technological foundation and expert support,” said Carlos Aravechia, head of Data Design & Intelligence at Numen.
The success of the project has validated a broader conviction at Natura &Co: that generative AI, embedded directly in ERP workflows, can fundamentally reposition finance from a transactional function to a strategic business partner.
A blueprint for other businesses
The Natura &Co project demonstrates a pattern that other organizations can replicate, particularly those running SAP S/4HANA. The combination of structured ERP data with the contextual reasoning capabilities of large language models creates a foundation for decision intelligence that goes well beyond traditional business intelligence tools.
The project was built within a six-month co-innovation sprint and went live in August 2025. It is currently in use across Natura &Co’s Equador operations.
Looking ahead, Natura &Co is already planning the next phase: integrating Joule Agents to further automate the extraction of standard analytical content and deepen the AI-driven optimization of financial processes.
“The success of this initiative validates the transformative potential of embedded AI within our ERP,” Dias Garcia noted. “We are now ready to move forward—deepening these insights and integrating the capability of Joule Agents to maximize the extraction of standard content and further optimize our business decisions.”
For SAP customers evaluating how to move from AI experimentation to AI in production, the Natura &Co project offers a concrete, replicable model: start with a high-value, well-defined business process, embed AI directly into existing workflows, and build in human oversight from the start.
A girl from a disadvantaged rural family in central China topped this year’s gaokao, attracting numerous live-streamers eager to finance her education, which she declined.
The home of 18-year-old secondary school graduate Han Yaping in a Henan province village was recently bustling with live-streamers.
This attention came after Han achieved an impressive score of 699 out of 750 in the gaokao, China’s national college entrance exam.
She has received offers from China’s two leading universities, Tsinghua University and Peking University.
Han’s accomplishment is particularly remarkable given her family’s impoverished circumstances.
Her mother suffers from ankylosing spondylitis, an inflammatory arthritis affecting the spine, preventing her from working. Her father, who earns a living through farming and odd jobs, serves as the family’s sole provider. Han also has a younger sister.
The UK’s Financial Conduct Authority (FCA) published a landmark review on Monday that proposes recommendations to regulate the impact of artificial intelligence (AI) on the financial decisions made by consumers.
The review, titled the Mills Review, anticipates that both consumers and firms will start delegating “more financial decision-making to AI systems,” including for agreements, initiating transactions, and executing decisions “within agreed parameters.” One of the key findings of the review outlined that while AI can help bridge advice gaps and “support growth,” there remain risks “associated with fraud, cyber security, and consumer harm.” Conducting the review, Sheldon Mills highlighted that “AI can also amplify risks: bias, discrimination, exclusion, opaque decision-making (particularly when multiple AI models interact), misleading or hallucinatory advice and erosion of consumer trust.”
The review stated that presently, one in five adults in the UK are “already open to AI making decisions for them,” particularly when decisions feel “complex or high stakes.” It found that roughly 26 percent of the population “trust general-purpose tools such as ChatGPT, Claude or Gemini for financial advice” with little awareness that such platforms provide no “formal routes to recourse” or protections.
Overall, the Mills Review identified four areas that it anticipates will be impacted by AI in the financial sector: “the transformation of firms,” “new consumer journeys,” “a reshaped competition landscape,” and “amplified financial crime and cyber risk.” The FCA projected the shift in how consumers and firms consult AI to take place by 2030.
The Mills Review put forth seven “priority” recommendations to be considered by the FCA Board. It recommended that any transitions to autonomous AI models be monitored and that regulatory frameworks and perimeters be adapted and secured. The review called for the strengthening of “system-wide coordination and oversight,” the scaling up of the FCA’s AI Lab to enable it to support AI models and innovation for agentic finance, and an “AI-enabled agentic supervisory model” to be built and adopted. Finally, it recommended that a trusted “public-interest AI-enabled financial capability service” be developed.
The FCA announced, in the press release, that it will launch an AI “good and poor practice publication” in late 2026.
-
San Diego, CA43 seconds agoSan Diego Humane Society Releases 4 rare western spotted skunks into the wild
-
Milwaukee, WI6 minutes ago
Former Judge Hannah Dugan fined $5,000, won’t serve prison time, judge rules
-
Atlanta, GA13 minutes agoJermaine Dupri sues Sony Music over alleged $18 million royalty dispute involving So So Def artists
-
Minneapolis, MN16 minutes agoTraffic shift starts Wednesday on I-394 in Minneapolis
-
Indianapolis, IN21 minutes agoDC BLOX cuts building from data center plans near Irvington, makes environmental pledges
-
Pittsburg, PA28 minutes agoWegmans to start building store in Cranberry this month
-
Augusta, GA31 minutes agoAugusta, GA Has a Shortage of Homes for Middle-Income Earners
-
Washington, D.C36 minutes agoReflecting Pool being drained – again – as Trump administration tries once more to fix DC landmark – WTOP News
