North Shore Financial

The advisory team at North Shore Financial struggled with information overload and time-consuming report generation. Each financial advisor had to manually gather data from multiple sources – internal research reports, client portfolio data from spreadsheets, market news from financial websites – to prepare for client meetings or to answer ad-hoc client inquiries. For instance, if a client asked, “How did my portfolio perform this quarter and what changes do you suggest?”, the advisor might spend hours pulling performance analytics, recent market trends, and relevant research insights before formulating a response. This process was slow and could delay client communications. Moreover, staying on top of compliance was a constant burden; advisors needed to ensure their advice and reports reflected the latest regulatory guidelines, which meant combing through compliance memos whenever rules were updated. The firm wanted a solution that could streamline data retrieval and report drafting, ensuring advisors always had up-to-date information and could respond to clients quickly with well-founded advice.

BotHire developed a Wealth Advisor Assistant AI for North Shore Financial – a bespoke AI agent system to automate data gathering, analysis, and even draft communications. The solution acted as an always-on financial research aide for the advisors. When an advisor needed a client portfolio update or an analysis of a market event, they could query the AI assistant (via a chat interface or a report request form). BotHire’s system would then orchestrate a series of behind-the-scenes actions: retrieving the client’s latest portfolio metrics and transaction data, pulling in current market data (prices, indexes, relevant news headlines), and retrieving any pertinent internal research (for example, if the firm’s analysts recently wrote about the tech sector and the client holds tech stocks). With this retrieved context, the AI’s generative model would compose a draft response or report – for example, a personalized email to the client summarizing portfolio performance, explaining how a new interest rate change might impact them, and even suggesting portfolio adjustments. By leveraging RAG in finance, the AI could dynamically incorporate the latest financial information and the firm’s proprietary insights into its output, ensuring the advice was both up-to-date and tailored. BotHire’s solution essentially gave each advisor a virtual analyst who works 24/7, crunching numbers and writing first drafts of communications. Advisors remained in control – they would review the AI’s suggestions – but the heavy lifting of data synthesis and initial drafting was offloaded. This meant clients got faster responses and more frequent proactive updates.

North Shore Financials' AI assistant was implemented with a focus on data orchestration and tool-use by agents to mirror the workflows of a financial advisory team. Key implementation aspects included:

• Unified Data Retrieval Layer: BotHire connected the AI system to all necessary data sources. Client portfolio data was accessible via a secure API from the firm’s portfolio management software (e.g., positions, returns, risk metrics updated daily). Market data and news were fetched through integrations with financial data providers – essentially giving the AI real-time access to stock prices, economic indicators, and news feeds. All internal research reports and investment strategy memos were ingested into a vector store knowledge base, similar to the legal case. When a query came in, a Data Retrieval Agent would automatically pull the latest figures (like quarterly performance numbers) and run semantic searches on the research database for any relevant commentary (e.g., “interest rates” or “tech sector outlook”). This ensured that the generative step always had pertinent, up-to-date facts on hand​hatchworks.com​hatchworks.com.
• Agent Orchestration & Tool Use: The AI assistant was designed as a composition of sub-agents each handling a part of the task, coordinated by an orchestrator. For example, upon a request for a portfolio review, the orchestrator would invoke: 1) a Portfolio Analysis Agent (which could run a Python tool or serverless function to calculate performance metrics from raw data), 2) a Research Finder Agent (to retrieve relevant internal analysis or external news via RAG), and 3) a Report Generator Agent (an LLM that takes the compiled data and writes a narrative). These agents used a tool-like interface; the Portfolio Analysis Agent essentially had a “calculator” ability to run code for analytics (hosted via a cloud function), while the Research Agent used a semantic search tool. The LLM-driven Report Generator acted last, using a carefully crafted prompt that embedded the numeric results and snippets of retrieved research into the context. For instance, it might receive a prompt: “Draft a client-friendly summary of portfolio performance. The portfolio return is X%, outperforming by Y%. Include insight about the tech stocks – see the analyst comment: [insert excerpt] – and conclude with a recommendation (e.g., rebalance 5% from bonds to stocks).” The model’s output would then be a well-structured draft email or report ready for the advisor’s review.
• Serverless Automation & Scheduling: In addition to on-demand queries, BotHire configured scheduled triggers for routine tasks. A cloud scheduler (akin to a cron job) would activate the system monthly to generate Portfolio Review Reports for each client automatically. This workflow ran off-hours, pulling in end-of-month data and any recent research notes, so that advisors would arrive each month to draft reports already waiting in their inbox for finetuning. Similarly, regulatory updates were handled by a scheduled Compliance Monitor Agent – whenever regulators released new guidelines or laws, a serverless script added those to the knowledge base and triggered the AI to summarize “What’s changed?” for the team. This proactive flow meant advisors were always briefed on compliance changes without having to manually read through dense regulatory text.
• Parallel Processing & Prompt Chaining: To keep responses quick, many retrieval and analysis steps were done in parallel. For a complex query (e.g., “Given the recent Federal Reserve announcement, how should we adjust Client X’s portfolio?”), the system split tasks – one thread fetched macroeconomic data about the Fed announcement, another pulled Client X’s bond holdings data, another retrieved the firm’s commentary on interest rate impacts. These results were merged for the LLM to generate a coherent answer. The orchestrator managed these parallel chains and handled timeouts or errors gracefully (for instance, if an external data API was slow, the system would proceed with what information was available and note any missing pieces). The design was influenced by the idea of prompt chaining in advanced AI agent setups​, ensuring that intermediate results (like “interest rates rose by 0.5%”) fed logically into the final answer’s prompt for the LLM.
• Human-in-the-Loop and Interface: North Shore’s advisors accessed the assistant via a secure web dashboard and an email integration. Through the dashboard, they could type questions in natural language (“Brief me on Client Y’s portfolio changes this week”) or select from templates (like “Generate quarterly report for Client Z”). The AI would present the draft answer or document in seconds. Advisors could provide feedback with a simple interface – e.g., clicking “regenerate with more detail on section X” or editing the text directly. The system learned from these preferences over time, subtly refining its prompts and retrieval priorities based on what advisors edited out or kept. This human-in-loop approach ensured the final output remained under expert control, vital for both quality and compliance.

North Shore Financial’s advisors quickly felt the benefits. Response times to client inquiries shrank dramatically – what might have taken a day or two to research and formulate (e.g. a custom email about a market event’s impact on a portfolio) could now often be turned around the same afternoon. Advisors estimated that about 30-40% of their weekly workload (in terms of report prep and data gathering) was eliminated by the AI assistant. In concrete terms, this meant each advisor saved around 8-10 hours per week, which they could reallocate to direct client interactions and strategic planning. The firm also noted that the consistency and depth of analysis in client communications improved. The AI’s ability to draw on the full breadth of the firm’s research meant no relevant insight was overlooked in client updates. For example, if the firm had a niche insight on tech stocks, the AI would invariably include it for clients with tech holdings, whereas an individual advisor might have missed that report in the past. This led to more informed clients and increased confidence in the firm’s advice. In fact, client satisfaction scores rose in post-meeting surveys – clients frequently commented on the promptness and thoroughness of the information they received.

Internally, the system also strengthened compliance and knowledge retention. Advisors were automatically alerted to regulatory changes with clear summaries, reducing the risk of compliance slip-ups. By having an AI that “remembers” all past advice and data, the firm created a sort of institutional memory – even if a particular advisor left, their successors could easily query the AI for historical context and rationale behind past decisions. Moreover, the cost savings were notable: the firm avoided hiring several additional junior analysts that they initially thought they needed to handle growth. Instead, the AI assistant scaled with their client base. As a side effect, junior staff who were already at the firm learned faster – by seeing the AI’s drafts and analysis, they got up to speed with the firm’s style and knowledge more quickly (almost like an interactive training tool). Overall, BotHire’s orchestrated AI solution gave North Shore Financial a cutting-edge efficiency boost, allowing a mid-sized firm to punch above its weight in personalized client service, much like how larger institutions leverage technology to augment their services.