Which company is this case about?
Morgan Stanley is one of the world's leading financial services firms, engaged in investment banking, securities, and wealth management. Particularly in its wealth management division, financial advisors with specialized knowledge provide services to clients, and maintaining and improving the quality of their work is constantly required.
Challenge to solve
Wealth-management advisors needed to find relevant information across a growing internal corpus and verify it in a form useful for client work. The previous figure of 70,000 reports per year is removed because it conflated this Assistant with a separate Research product.
Morgan Stanley Wealth Management worked with OpenAI on AI @ Morgan Stanley Assistant, an internal tool that retrieves firm knowledge to answer advisors' questions. OpenAI says the corpus expanded to 100,000 documents. AskResearchGPT is a separate Morgan Stanley Research product.
- ▸More than 98% of wealth-management advisor teams use the Assistant.
- ▸The share of documents accessible through the system increased from 20% to 80%, according to OpenAI.
- ▸GPT-4, retrieval work, evaluations, and zero data retention are confirmed; the detailed internal data flow is not public.
How AI was used
Morgan Stanley Wealth Management worked with OpenAI on AI @ Morgan Stanley Assistant, which retrieves internal knowledge to answer advisors' questions. OpenAI says the corpus expanded to 100,000 documents.
GPT-4, retrieval work, evaluations, and zero data retention are public. The indexing, ranking, prompt, and application details are not public.
Implementation Effects and Key Takeaways
- ▸Confirmed: GPT-4 supports the internal Assistant and OpenAI describes work on retrieval methods and evaluations.
- ▸Confirmed: OpenAI provides zero data retention for this deployment.
- ▸Not public: the detailed indexing, ranking, prompt, and application architecture. These components should not be presented as confirmed.
What Japanese companies can learn
This case study is applicable not only to the financial industry but also to any industry with a large volume of specialized knowledge and internal manuals, such as legal, medical, and manufacturing. Enabling everyone to utilize the 'knowledge assets' accumulated within the company through AI directly leads to improved employee productivity and service quality. An effective approach would be to initially introduce it experimentally as an FAQ response system for a specific department or an internal policy search system, and then expand it while measuring its effects.
A separate product
AskResearchGPT is a separate Morgan Stanley Research product and should not be conflated with AI @ Morgan Stanley Assistant.