Artificial intelligence (AI) and machine learning are reshaping how financial markets process information, identify potential outperformance and manage risk. This evolution is also opening new possibilities for index construction. By incorporating machine-learning-driven alpha forecasts into transparent, rules-based methodologies, indices can be constructed using vast and complex datasets, supporting more adaptive, data-driven approaches.
S&P Dow Jones Indices (S&P DJI) is helping advance this evolution through the launch of the S&P 500 3AI Top 100 Index and S&P World 3AI Top 300 Index. The S&P 3AI Indices leverage 3AI’s Alpha Intelligence scores (3AI Scores) within S&P DJI’s established index framework to identify companies that exhibit high forecasted excess return potential, providing a systematic, AI-enabled approach to stock selection.
This FAQ provides an overview of the S&P 3AI Indices, including their methodology, the role of 3AI’s machine-learning-driven forecasts, validation and governance processes, back-tested performance characteristics and their potential relevance for market participants.
1. How are the S&P 3AI Indices constructed? The S&P 3AI Indices track companies within their respective underlying universes that exhibit the highest forecasted alpha, as measured by 3AI Scores. The methodology for the S&P 500 3AI Top 100 Index selects the top 100 companies from the S&P 500, while the S&P World 3AI Top 300 Index methodology selects the top 300 companies within the S&P World Index.
At each quarterly rebalance, companies are ranked by their 3AI Scores, and the highest-ranked companies are selected until the target constituent count is reached. Selected constituents are then weighted in proportion to their 3AI Scores, assigning higher weights to companies with stronger forecasted alpha, subject to a 40% GICS® sector cap and a 4.5% single-stock cap.
2. Who is 3AI? 3AI is a London-based, AI-driven investment intelligence company founded in 2018. Its team combines decades of institutional asset management experience with deep expertise in machine learning and applied data science. 3AI has received multiple industry awards, including becoming the first AI firm to win “Trading System of the Year” at the 2025 Futures & Options World International Awards.
3AI specializes in applying machine learning to generate systematic alpha forecasts, which are supplied to several tier-one hedge funds, systematic asset managers and institutions. 3AI’s AI technology produces forecasts, known as 3AI Alpha Intelligence Scores, for over 20,000 stocks globally. S&P DJI has a commercial agreement with 3AI that allows S&P DJI to develop, calculate and license indices that leverage 3AI Scores.
3. What is machine learning and what role can it play when incorporated into index construction? Machine learning is a branch of AI in which computers learn from data and make predictions or decisions without being explicitly programmed for a specific task. Rather than relying solely on pre-defined rules, machine learning models can learn from large and diverse datasets, adapting as new information becomes available.
When incorporated into index construction, machine learning may enhance the ability to process and interpret complex data at scale. These may include fundamental, technical, macroeconomic, sentiment or alternative data that may be difficult to evaluate using traditional rules-based approaches alone. Machine learning can also help identify relationships across variables, detect non-linear patterns and generate systematic forecasts, such as expected return, risk or alpha.
For indices, these capabilities may support a more dynamic and data-driven approach to constituent selection and weighting. However, robust governance, transparency, model validation and risk controls remain essential to help ensure the approach is systematic, repeatable and appropriate for index use.
4. What are the 3AI Alpha Intelligence Scores? The 3AI Scores are proprietary forecasts generated by 3AI that provide a 12-month excess return forecast for each company relative to a global universe of approximately 20,000 stocks.
The scores are produced through a fully systematic, machine-learning-based process supported by validation checks and oversight procedures. S&P DJI incorporates 3AI Scores into the index construction process to help select constituents and determine weightings within the S&P 3AI Indices.
5. How are the 3AI Alpha Intelligence Scores generated? The 3AI Scores are generated using a series of machine learning models that process a broad range of data to identify patterns and produce forecasts. The scores are updated weekly using newly available data, allowing the forecasts to evolve over time as the underlying information set expands.
The final 3AI Score combines two scores. The first is a bottom-up single stock score, which analyzes hundreds of company-level data, including fundamental data, technical indicators, factors and other alternative datasets. The second is a top-down business cycle score, which analyzes macroeconomic and sector-level data. The final score is calculated as the simple average of these two scores.
To improve reliability and help manage model risk, the combined score is further refined through a Bayesian Believability Layer. This layer applies statistical techniques to adjust for potential overconfidence in the model outputs and to account for uncertainty, serving as an additional validation and risk-control mechanism.
