SSF Pilot - Extracting insights from Stewardship reports

Context

"Stewardship" is a sustainable investment approach that refers to the owner of a company’s stock engageing with that company in order to influence its corporate policy towards more sustainable business practices.

Investor stewardship can be implemented through active voting at the company’s annual shareholder meetings and through engagement and dialogue with the company’s board of directors, C-level management and other senior representatives. Concerted voting and engagement can lead to improved sustainability strategies and processes and – as a consequence -to improved ESG performance and reduced risks. If executed diligently and with a sound business understanding, stewardship has the potential to drive meaningful change and the smooth transition towards a sustainable economy.

In the past few years, Stewardship has established itself as one of the main sustainable investment approaches applied by financial services companies in Switzerland (see SSF’s Sustainable Investment Market Study 2025, page 36, Figure 22).

A significant portion of Asset Managers (more than 40%) and a great majority of Asset Owners (83%) do not perform stewardship by themselves, but rely on a third-party provider – another portfolio manager to which the active management of one’s portfolio has been delegated, or a specialized service provider on the topic (see SSF’s Sustainable Investment Market Study 2025, page 39, Figure 26).

Challenge to overcome

In both cases, this means that a large number of Asset Managers and Asset Owners rely on third party to execute stewardship activities on their behalf, and on their reporting to assess the effectiveness and these activities. Most service providers wil not provide a customized reporting to each investor who entrusted them with the execution of stewardship activities. Investors who have delegated stewardship execution need to browse the stewardship and other reports from the service providers in order to find the information they need. If they have several providers, they need to screen several Stewardship reports in different formats - the workload increases accordingly, which, especially for some Asset Owners with limited resources, can become an issue.

AI application / Use case

At SSF, we have developed a pilot tool to help Asset Owners (and over investors who delegated portfolio management and stewardship execution) navigate such reports much quicker and extract the key information they need in seconds.

This use case relies on "Intelligent Process Automation", i.e. the application of AI capabilities to the execution of recurring tasks on datasets which are similar but not identical – in that case reports that tackle the same topic (stewardship) but with different formats, structures, narratives, levels of detail, etc.

The tool acts on behalf of the Asset Owner / investor and asks questions destined to the service provider on the execution, progress and effectiveness of stewardship activities like a human employee or the investor themselves would do. It then navigates available information from the service provider reports, identifies and extracts relevant information, and populates the response in a standardized questionnaire.

1) Questionnaire before completion - questions are prefilled, can be adapted at will

2) Script is being run - extracts questions and send them as prompt to the LLM, LLM returns responses based on content of the stewardship report analyzed (PDF)

3) Questionnaire completed with the answers extracted by the LLM from the stewardship report of the asset manager

The Asset Owner / investor then only has to navigate the prepopulated responses and perform a quality check on the relevance of information gathered, the level of detail obtained, etc.

You can download below one example, based on Blackrock's Investment Stewardship Annual report 2024:

• BlackRock's report
• Excel questionnaire filled by our pilot tool (no human modification was done to the output)

Benefit to the financial services industry

Teams covering sustainability at Asset Owners are often under lots of time pressure. The application of this tool can greatly improve the efficiency of data gathering for Asset Owners on stewardship activities done on their behalf - thereby releasing time sustainability teams to perform the other tasks due.

Technical solution

• We wrote a Python script that extracts predefined questions from a Excel questionnaire (questionnaire downloadable here) and sends them to the Large Language Model GPT-4o via the OpenAI’s Application Programming Interface (API)

• The Python script also extracts and chunks via the OpenAI’s API the data from a PDF file that is submitted to it in a predefined folder (e.g. an Asset Manager stewardship report). It leverages for this a number of Python libraries like numpy, tiktoken, nltk, PyMuPDF…

• The LLM processes the questions based on the data from the PDF report, send back the responses through the API

• The Python script records the response in Excel and creates a new file "questionnaire_[assetmanagername]_filled.xlsx"

Benefits of the use case to the financial services sector

This use case is a simple pilot and further work is required to make it a usable tool for Asset Owners and investors. However, if scaled, that tool would allow significant time spare and efficiency increased on the side of Asset Owners, freeing up resources for other tasks.