Types of Artificial Intelligence
Artificial Intelligence (AI) is an umbrella term for many different technologies including machine learning, neural networks, expert systems and large language models (LLM). For many years RightBRIDGE has been using expert systems to provide a rules-based approach to provide firms with automated “guardrails” to adhere to regulatory requirements. Rule-bases provide a consistent and repeatable approach by applying consistent rules and parameters to reinforce compliance standards. From a compliance standpoint a user must “show their work”. The ReasonText™ provided by RightBRIDGE is used to describe the analysis and results. CapitalROCK is now introducing RightBRIDGE+ which includes additional functionality using large language models otherwise known as generative AI. RightBRIDGE+ uses the Azure Open AI (also known as ChatGPT) to provide added functionality to supplement the battle tested rule-based system currently deployed in the RightBRIDGE solution. An important note, the large language model is only applied to the existing content within the 4 walls of the RightBRIDGE environment to limit the analysis to reliable reference material.
Current RightBRIDGE+ Features
RightBRIDGE AI Note Assistant
One of the largest causes of NIGOs is insufficient or inaccurate case notes provided by financial professionals documenting the details and recommendations of products for suitability and regulatory compliance requirements. The AI Note Assistant is designed to enhance the review process of financial professional submitted case notes written by the financial professional. ItThe RightBRIDGE+ Note Assistant will analyzese the user’s submittedfinancial professionals note notes wiin context withth the the corresponding ReasonText™️ from the relevant RightBRIDGE application. section and aTo give direction to the LLM a configurable system message, or prompt, is sent along with the note and the ReasonText™. using Azure’s instance of the OpenAI Large Language Models (LLMs).
ReasonText™️ for a particular item is combined with the user note and the firm configured instructions (prompt) into the Large Language Model (LLM). The LLM returns to the system a response that indicates what has been identified as missing in the note.
How it Works
Upcoming RightBRIDGE+ Features
Case Summary
RightBRIDGE is currently implementing AI functionality using an LLM (Large Language Model), Generative AI to summarize the case data down to a format that will provide users a simplified view of a case. This is a configurable output based on the client’s needs. Data and results including ReasonText™ and scores are passed to the LLM. A prompt that is configurable by the firm is analyzed with the data. The resulting output can be formatted as a summary report or displayed on screen. The following is an actual summary generated by the AI solution. This sample is based on a report that is ~20 pages. Full-length RightBRIDGE reports are also transmitted with the Case Summary.
Sample paragraph
Patrick Childs’ is rolling over $300,000 from a 401k into an annuity. Patrick Child’s has $680,000 in investable assets. Patrick desires a balance between guarantees and liquidity, would like to manage their own assets, and does not have a strong opinion regarding fees. The main purpose for the annuity purchase is for Income/Living Benefit Riders. The initial withdrawal is $22,275 annually, which lines up with Patrick desire for income through the purchase of this annuity. Patrick has a moderate aggressive risk tolerance. It is worth noting that the total cost of the proposed variable annuity will be 1.50% annually.
AI-powered Document Reading
Currently within RightBRIDGE, users are asked to input information from customers related to fees on accounts or current holdings. For 401(k)s and other ERISA accounts the best source of this information is in the customers 404(a)(5) fee disclosure. A secondary source could be an account statement. Fee disclosures and statements can vary by plan sponsor so there is no consistent process for the financial representative when searching through these documents and there is the possibility of error as there is with all manual processes.
The LLM-powered document reading function streamlines the manual process of filling out forms by allowing users to upload documents (such as a 401(k) plan statement, annuity prospectus, or client statement) within the RightBRIDGE application. The AI reads and extracts relevant information from the document, such as names, fees, and other key details, and automatically populates the necessary fields in RightBRIDGE. There is an intermediary step that allows the user to confirm the details pulled from the document were pulled as expected. This ensures faster, more accurate data entry and reduces the risk of human or AI error.
How does this function work? The Financial Representative uploads documents into the application. The document(s) is then passed to the large language model (LLM) with a prompt instructing the model to pull the key information needed. The application will then provide a response to the financial representative with the extracted data. The representative will have the option to select the data to use (human in the loop). Once selected this data will populate the RightBRIDGE questionnaire.
