The 2024 AI Maturity Index is composed of the top 200 companies listed in the 2023 Fortune 500 list. The scoring for each company is a summation of the resulting score in each of the three components below:
1. Content similarity analysis
Machine learning algorithms were used to analyze the content similarity of corporate texts, which included 2022 corporate annual reports, 2023-2024 analyst quarterly calls, and 2023 company press releases, against an AI benchmark book, HBR’s 10 Must Reads on AI, published in 2023, selected due to its popularity and its broader approach to AI transformation issues.
Two types of similarities were measured: lexical similarity (which captures the number of words in common) and semantic similarity (which captures the similarity of the meanings of those words). To measure both types of similarities, a combination of cosine similarity (which accounts for lexical similarity) and text embedding (which accounts for semantic similarity) were used.
An analysis was conducted on each of the three data sources – annual reports, analyst quarterly calls, and company press releases. The average of the three data sets was used as the final score.
2. Number of AI-related keywords
A list of AI-related keywords was developed via desk research and with categorization algorithms. These keywords were used to analyze two AI-related books: Artificial Intelligence for Business and AI Fundamentals for Business Leaders.
This process generated 16 distinct categories, with each category consisting of approximately five keywords. Each company received a score for each of the three data sources ranging between 0-16, depending on how many categories of the 16 AI categories were mentioned. The scores from each data source were standardized for each company. The final score was calculated as the average of the Z-scores from each data source.
3. AI deals and amounts invested
Using data from Pitchbook, the number of AI-related deals from 2019 to September 2024 was collected. Additionally, the investment amounts (in million US dollars) for each company, during the same period, were also recorded.
Both AI-related deals and investments were normalized using the ln(𝑛+1) transformation. Each variable – the number of deals and the investment amounts – was then standardized separately and averaged to produce a final score.
4. Content similarity
This component assesses whether a company talks about AI the way an organization that understands AI transformation would. Corporate texts from the same three sources were compared with a benchmark text, HBR’s 10 Must Reads on Artificial Intelligence, Updated and Expanded (Harvard Business Review, 2026), chosen as a widely read, practitioner-oriented synthesis of AI transformation.
Two types of similarity were measured: lexical similarity, the cosine similarity between word-frequency vectors, which captures the words texts have in common; and semantic similarity, based on text embeddings, which captures shared meaning even when the wording differs. Each source was analyzed separately, and the average of the three was used as the component score.
Weighting and final score
Each component is standardized so that measures in very different units can be combined. The overall score is their weighted sum: 30% each for AI deals and investment and AI patents, 25% for keyword coverage and depth, and 15% for content similarity. Companies are ranked on that score.
The weights follow a simple principle: what a company does counts for more than what it says. Investment and patents require money, people and time, so together they account for 60% of the score, against 40% for corporate communications.
The ranking is not sensitive to the exact choice of weights. With all four components weighted equally, the rank correlation with the published ranking is 0.99.
Notes on interpretation
Comparisons with the 2025 edition should be made with care. Between the two editions the sample grew from 300 to 330 companies, the patent component and new weights were introduced, the keyword list was revised and the investment window moved, so rank changes combine genuine change with methodological effects.