Mo, the First Personified Next-Gen Chatbot for Investment Research
Investment research continues to see increased deployment of ChatGPT-like capabilities both from large incumbents and from startups.
In my early May blog ChatGPT Empowering Investment Research: Trends and Use Cases, I mentioned 3 incumbents and 4 Fintechs. The list is growing as we speak.
Today, my focus is on the first personified advanced AI chatbot in investment research. Meet Mo, the Morningstar creation.
Unlike David Roeder who covers business and labor for the Chicago Sun-Times[1], I didn’t get to ask Mo, what the market outlook is for 2023, or what money moves should I consider if I believe that rates will continue to rise in the US, or whether the stock market is a ponzi scheme.
However, I did get to speak to Ravi Sarkar, Enterprise CTO focused on Capital Markets and Fintech at Microsoft, to learn more about Mo who is trained based on the know-how of Microsoft’s Azure OpenAI Services but limited to Morningstar’s sizable investment knowledge base.
Mo, is in beta testing across Morningstar’s flagship products Morningstar Investor, Research Portal, Direct, and Advisor Workstation platforms. Morningstar Investor is a market-leading independent research, ratings, and tools for individual investors. Morningstar Direct is an investment research software that lets investors compare investments and portfolios and analyze market trends. The annual Morningstar subscription for individual investors is $249 and includes access to Mo. The rest of the Morningstar products, Morningstar Research Portal and Morningstar Advisor Workstation, are mostly for financial advisors.
Keep focused on your domain expertise
First and foremost, Mo, is well-trained to say I don’t know if asked questions that are unrelated to investments. Ravi Sarkar emphasized the importance of extensive negative testing in training Mo. These chatbots are prone to hallucinations, so making sure that Mo provides consistently a certain answer in every possible negative scenario, is paramount.
Learning by interacting but what about privacy
These Transformer LLMs that are powering up OpenAI, do learn from our interactions but that does not mean that the conversational interaction and the insights the user requests through a prompt belong to the knowledge library. So, if I ask Mo for example `Which Thematic investment product category outperformed in 2022 by annual performance and also based on the Assets attracted?` the response provided will not be included in the Morningstar library. In that sense, there is no feedback loop. The Morningstar library is NOT augmented with the new information and nobody can access my question and conversation with Mo; but the language model is learning from this conversation.
In terms of privacy and security of the conversations with Mo, there are two levels of control in place. The first is, a user`s access level to Morningstar`s services will determine the access and the type of information that Mo provides to the user. Any individual user interactions are not available to any other person.
The second guardrail is that all the information exchanged between the user and Mo, is residing within the customer’s tenant, within Morningstar’s Azure tenant, which is encrypted.
How can Mo, delight a user?
Here is a concrete and simple example that shows the power of an AI-chatbot like Mo.
I posed a question to my traditional search engine: `Give me the 10-year P/E ratio adjusted for inflation for Microsoft`
No matter which stock I chose, I got a list of links to pages on these websites: ycharts.com, CorporateFinanceInstitute, Wikipedia, etc.
Mo, on the other hand, can provide a very concise and fast answer. This is a major upgrade in how machines can provide information in an intelligent, concise fashion, especially when trained in domain-specific data.
We will be seeing more training on large financial knowledge libraries, like those of Morningstar. Evidently, businesses that have large proprietary knowledge libraries that lend themselves to conversational interactions (unstructured data) will afford to train such Transformer models and offer valuable services to their clients.
Conclusion & reflections
Mo, first appeared publicly in April at the Morningstar Investment Conference and conversed with Morningstar CEO Kunal Kapoor on stage. Attendees got to ask Mo, thousands of questions, as he tirelessly answered.
For now, Mo is a value-add in beta testing mode to existing Morningstar services and to paying subscribers. There is no access to Mo via API or via some plugin. However, with this extremely fast pace of pilot launches that leverage the advanced LLM capabilities, the customer experience bar is rising fast. End customers, individuals, and businesses are starting to expect these capabilities.
Investment Research is one of the first fields that will largely benefit from these integrations. This is an area that I am following and will be launching soon an interactive tracker map of this space.
This is a sponsored post. Microsoft is my client. I choose to highlight topics that I believe are valuable in understanding how technology is reshaping Financial Services.
[1] Morningstar’s AI chatbot gets investors closer to mostly right answers https://buff.ly/45DYoSO
📌 Twitter: https://twitter.com/efipm
📌 Subscribe to my YouTube Channel with my insights and industry leader interviews. New video every Wednesday: https://www.youtube.com/EfiPylarinou
📌 Spotify Podcasts. Follow here: https://open.spotify.com/show/5bRkZEYHSwPiGx7vTqylw6?si=Mg3hN5PDQ86K10GjeK52jw
📌 Linkedin: https://www.linkedin.com/in/efipylarinou/
📌 TikTok: https://www.tiktok.com/@efiglobal
📌 Web: https://efipylarinou.com/
