Sustainable Investor AI: GenAI-based advisor for ESG Investments

GOAL

AI advisor for ESG (Environmental, Social, Governance) investments

PROBLEM

• 86% of millennials consider ESG important
• Quantity of global assets managed according to sustainability investment strategies more than doubled from 2012 to 2018, rising from $13.3 trillion to $30.7 trillion.
• Investors value sustainability information but complain about: i) data quality ii) greenwashing iii) consistency and reliability
• Manual analysis of reports is time consuming due to volume & complexity of data

SOLUTION

Safe investment AI advisor across traditional financial factors & ESG metrics; leveraging GenAI for research and measuring trustworthiness of data.

Conversational Interface

• Leverage GenAI to summarize, analyze financial research & sustainability reports

Use GenAI to measure trustworthiness against greenwashing

• Spot greenwashing against industry standards
• Embed Trustworthiness score in results & links to sources for Investors.

HYPOTHESIS

Leveraging GenAI & embedding results with the trustworthiness of ESG data sets, we can provide safer/more reliable financial research advice in an easy to use natural language.

PROMPT TUNING TECHNIQUES USED

Role based prompting.

Excerpt referencing.

Focused & separate questions.

Including company metadata in the prompt.

Specific criteria for quality evaluation scoring.

Clear & specific instructions.

RESULTS

Choose LLM model per use case.

Need for prompt guidelines, clear instructions, be specific & concise.

Traceability improves reliability: asking LLM to tell me which sentences it uses to generate the answer improves results immensely.

Tradeoff between predictability & creativity: Adjusted temperature, Trustability check required greater predictability.

Domain specific prompt development: Important to have domain expert involved in the prompt engineering.

Team Members

• Anshul Madan (madananshul@gmail.com)
• Aycan Katitas (aycan.katitas@gmail.com)
• Akash Reddy (akash.r7382@gmail.com)
• Mengdi Huang (mengdih6@gmail.com)

FEEDBACK

Seeking feedback on Sustainable Investor AI project, happy to hear feedback over email or even have a short 30 minute call (e.g. how useful you think this product could be, other adjacent areas we could look at, or ideas on datasets we could use)

Mail: Anshul Madan madananshul@gmail.com