AI supporting portfolio managers in sustainable investment decision-making
Challenge to overcome
The 17 Sustainable Development Goals (SDGs) were established in 2015 by the United Nations. They can be broken down into 169 actionable targets, with 232 indicators used to measure the progress toward each goal.
Investors are however confronted with a lack of a standardized reporting framework and the absence of common reference data across companies to estimate both the goals and the targets at company level. This leads to limited comparability between company-reported data and reservations on the reliability of such data.
In addition, most companies focus their SDG alignment reporting on their own operations, without capturing the impact of products or services that help others to develop towards a more sustainable future.
This drove Pictet to develop a systematic and independent tool to assess the SDG alignment of the companies composing their discretionary management portfolios, and to identify SDG improvers, i.e. companies that are on a transition journey.
AI application / use case
Pictet decided to develop their own AI-based model to overcome data challenges and establish a common reference framework for their investment managers and client reporting. Pictet’s tool relies on Natural Language Processing (NLP) and machine learning techniques which analyses various data sources to extract the company activities and measure their positive or negative alignment to each SDG. The tool consists of several steps:
1. The first step collects information from different sources (descriptions, website, financial reports, earnings call transcripts etc.) covering a variety of information about each company.
2. The second step extracts keywords for each company and measures the relevance within each data source. It then aggregates this data to form a company’s fingerprint of its most relevant keywords.
3. In the third step, the tool extracts the most important concepts from the 169 SDG sub-targets and classifies each concept positively or negatively to the contribution of the SDG. Pictet has engaged a third-party provider specialised in the analysis of scientific papers for this step and has developed a database of approx. 5000 keywords that are positively or negatively related to the SDGs.
4. Finally, as last step, the tool combines both the SDG concepts and company activities together to assess the positive or negative alignment of a company with the SDGs. This NLP-based assessment is combined with companies’ revenue data to further improve the outcome quality.
The model provides detailed positve and negative alignment scores for approx. 40’000 companies to each of the 17 SDGs. The tool can then calculate the average alignment of the portfolios and the reference index with each SDG. The final output for our clients is a chart demonstrating the strength of the relationship between the SDGs and the strategy relative to the global equity market (using MSCI ACWI as a proxy).
Figure 1: Illustrative example of the SDG profile of a Thematic strategy
Use case key beneficiaries
☐ Relationship Managers
☒ Portfolio Managers
☐ Research teams, macroeconomists
☐ Control functions
☐ Support functions (HR, CFO, …)
☒ Other: Clients
Supporting technology
The system leverages state-of-the-art Natural Language Processing (NLP) and advanced machine learning algorithms to assess and quantify the impact of companies on the United Nations Sustainable Development Goals (SDGs). By harnessing AI-driven keyword extraction and semantic similarity analysis, the tool intelligently interprets vast amounts of company data – including business descriptions, corporate documents, and detailed revenue streams.
This allows to deliver actionable insights and transparent reporting for sustainable asset management.
