We sit down with P. Venkatesh, Co-founder of Maveric Systems, to talk about how they are leveraging Chat GPT in the Asset and Wealth Management sector, data privacy in the AWM sector, and more.
Can you suggest some specific benefits ChatGPT brings over the traditional customer interaction techniques in the Asset and Wealth Management sector?
Investor profiles have been undergoing substantial changes in the last few years; this includes millennial ultra-high net worth investors, women who have inherited wealth, and people without financial literacy.
Traditional customer engagement platforms can be enhanced with ChatGPT to provide information, market data, and guidance as it has the ability to provide unambiguous, confident, and straightforward information.
Finance ChatGPT platforms facilitate investors with relevant data and guidance which can be conveyed to the investors in real-time. It improves in enhancing customer experience not just through personalized information which comes through contextual response, but also through consistent messaging on customer, product, and brand.
What are the growing concerns Maveric Systems witnessed or expects about data privacy and security while integrating ChatGPT into the operations of AWM enterprises?
Three primary concerns have been coming to the forefront –
Ultra-high net worth and high net worth investors like protecting their privacy. Typically, getting their consent for moving their personal data to public spaces is a tough challenge. Data of such investors need to be masked before using them for analysis, even if used internally. This makes it difficult to create data sets that can be used for ChatGPT, as the model functions well only when trained with large data sets.
In order to provide timely, accurate, and relevant information to the investors, integration with specialized platforms aligned to finance is needed. As most financial platforms hold public data, the integration needs to be secure. It is in a nascent stage now and the security of these platforms would only improve over time.
The model’s efficiency to search for relevant information is at a different level of maturity than what was predicted initially. The model is yet to be production ready.
As an expert, what key considerations will Maveric Systems suggest to AWM companies while evaluating the potential of ChatGPT in their business? Any specific use or example Maveric Systems would like to share?
While evaluating the potential of ChatGPT, AWM companies need to consider the following factors –
Data Quality – AWM requires maximum trust levels, even within banking. Therefore, it would be prudent to rely on internal data, and only on regulated external data that is reliable.
- Creating an instruction set that is suited to the purpose.
- Fine-tuning the model by testing the tokens and instruction set created over a period of time.
- Including relevant components needed to ensure data consistency, lineage, and explainability.
Some of the use cases which we consider relevant are:
- Vertical models
- Customer engagement with financial news
- Sectoral performance across markets and geographies
- XAI
- Customer engagement with personalized information and context-related response
- Consistent messaging on customer, product, and brand
- AML entity grouping for KYC, CDD, PEP, and sanctions check
- Chatbots
- Enhanced customer engagement
- Content aggregation and classification for querying and analysis
- Customer experience
- Customer engagement with personalized information and context-related response
- Consistent messaging on customer, product, and brand
According to Maveric, what are the key challenges in leveraging ChatGPT to scale asset and wealth management?
The key challenges are around –
- Building large volumes of reliable data, both financial and transactional.
- Model interpretability. While ChatGPT provides a response, it does not substantiate that with data or reasoning. In the financial segment, explainable components become important, and in some cases, it is a regulatory requirement too.
- ChatGPT operates only in the textual formal currently, while many social media platforms demand image, voice, and audio components to be included on the platform.
- Language interpreters may be needed in many markets where English is not the dominant language of choice for customer interaction.
Please help with the insights towards the potential scalability of Generative AI to cater to wealth management and also the limitations which come along.
Let us look at the developments that augur well for its adoption by the industry.
- Production grade components exist for most of the requirements, from language interpretation to model interpretability and conversion of audio, video, and image to text.
- Finance vertical ChatGPT versions are emerging with a large volume of data sets and instruction sets.
- While large volumes of internal and regulatory data are available in digital form, cleaning and improving the data quality is still an issue.
The scalability of large language platforms is proven. For example, BM Watson can hold 36 petabytes of data and ChatGPT can hold 175 billion, establishing their ability to scale.
The challenges that arise are around –
- Securing large data storage on public or private platforms; data breaches do occur, and sensitive customers would not prefer such breaches.
- With the components of audio, video, and images-to-text conversion, it is currently difficult to ensure privacy.
- Data quality would remain an issue; large-scale data cleaning can be erroneous.
What role does Maveric Systems play in leveraging the AI ecosystem into AWM for Fls?
Maveric engages with customers on the following –
Based on the given context of the bank we illustrate the value of some of the most relevant components like the vertical model, XAI, chatbots, and customer experience needed to leverage the AI ecosystem.
While outlining the key challenges in leveraging ChatGPT, we brought out the components that may be needed in certain contextual usage of the bank. We create those integrated components in our lab depending on the requirement, and then create or access public data or the bank’s data to show the model.
We actively work with the bank’s innovation team on their deep tech initiatives to establish the use cases and their value.
We also work on building the model right from the data sets, and instruction sets, fine-tuning those, and finally testing and certifying their quality.
According to a McKinsey report, by 2030 “80% of new wealth management clients will prefer advice in a Netflix-style model i.e., a hyper-personalized data-driven method. What are Maveric insights on this finding?
We believe that the following would become important in an engagement:
- Contextual historic data in a given market or across geographies.
- Contextual industry segment and index-related information in a given market or across geographies.
- Financial literacy on existing as well as new financial instruments, markets, sectors, and regulations.
- Customer engagement with personalized information and context-related response.
- Content aggregation and classification for querying and analysis.
Hyper-personalized advice would require not only ChatGPT but also the integration of other components like a vertical model, XAI, chatbots and customer experience to be built on top of them, along with an external integration to access the relevant data.
About the Author
As the Co-founder at Maveric, P Venkatesh (PV) leads the global core banking business that is now aligned with Temenos. PV and his solution architects successfully won 9 out of 10 engagements, having envisioned and launched industry-first test frameworks, automation, and contextual value-adds. A veteran of 30 years, consultant, and entrepreneur, across banking, financial services, retail, government, and urban services, PV’s deep competence in retail banking and regulatory compliance is sought across geographies.
The article was Originally published in Times Now.
