Which part of investment research has generative AI changed first?
- Discussions at the Bernstein conference suggested that AI offers greater incremental benefits to discretionary investors. Quantitative teams already made extensive use of automation and machine learning before generative AI became widespread, so AI adds more support in areas such as coding and information gathering. Discretionary research, by contrast, can use AI to lower the technical barriers to processing unstructured information.
- AlphaPai’s platform evolution reflects a shift in the focus of its applications: in 2023, it first turned information such as meetings into text and structured meeting notes; in 2024, it began covering company and industry research as well as analysts’ workflows; and in 2025, PaiWork brought data, workflows, research tools and AI agents together in one work environment. AI has gradually moved from standalone tools into everyday research workflows.
How widespread is actual use, and where is its value evident?
- AlphaPai says it serves more than 120 thousand professional users and covers about 8000 institutions. Over the past year, token usage on the platform grew 80-fold, but 5.5% of users consumed 50% of the tokens, indicating that usage intensity is uneven and activity is concentrated among a subset of users.
- Meeting bots turn content from investor presentations, expert discussions and similar events into searchable information, while ongoing-monitoring tools automatically update developments in companies, industries, technology and supply chains. They address time-consuming tasks involved in manually collecting, organizing and maintaining information, rather than directly replacing final investment decisions.
Can efficiency gains translate into investment alpha?
- Participants noted that some quantitative funds making intensive use of AI have performed strongly, but that does not allow their investment success to be attributed to AI. AI enables investors to test more signals; the more signals they test, the greater the risk of selecting a high-Sharpe-ratio strategy by chance, so backtested performance needs to be discounted accordingly.
- AI’s information-processing capabilities are more directly relevant to high-frequency investors. For medium-frequency investors, evidence that AI can improve decisions is limited. Historical data for medium-frequency macro strategies are scarce, making it harder to determine whether greater model complexity will lead to better future performance.
- AI is therefore better suited to information processing and efficiency tasks within research workflows. Investment professionals still need to assess, validate and improve the conclusions generated by AI.
What factors will determine whether AI can keep changing the investment industry?
- As AI takes on more information gathering and synthesis, people’s value shifts toward in-person research, expert interviews, independent judgment and quality control. Large platforms have advantages in scale and talent, but boutique firms can also use AI to broaden their research coverage.
- By processing new information faster, AI may enable markets to reflect information more quickly in the short term. If different investors use similar systems and reach similar conclusions, their holdings may also converge, bringing greater market volatility and periods of inefficiency. Participants also suggested that workflows relying less on AI could become a source of differentiation.
- The pace of adoption depends on whether tools can be integrated into everyday work and whether users can see clear productivity gains or investment returns. Evidence that AI improves investment decisions and performance remains an important hurdle.