Timeline Execution: Q2 2026
Industry: Investment Management with In-House Equity Research
Engagement: AI Advisory and Custom Build Work
The Problem
This regulated investment firm chose a deliberately careful Claude rollout. Staff use Claude chat, while desktop agent and coding tools stay off. The equity research team saw more potential, such as a morning check of covered stocks for big price moves, and the firm’s AI lead wanted the safest path to get there, within the firm’s AI policy and approved-application list.
The Solution
- Built a controlled data connection. A custom Claude connector (MCP) reaches a defined set of operations data through its own read-only service account, so Claude sees only what the firm chooses.
- Refined it with the client. The connector now tells Claude which objects are tables and which are views, and supports the database functions the team requested.
- Designed automation that keeps controls in place. A small scheduled runner in the firm’s own Azure environment reads a SharePoint task list, sends a fixed prompt and data to Claude and delivers results by email, Teams or SharePoint. The firm’s AI leads control the task list.
- Made small requests easy. One-off requests run as email-approved tickets with a set hour cap.
Did You Know?
AI agents can pull far more data than people do. By loading daily prices and dividend history into its own database, the firm expects to cut connector calls by about 90%.
The Results
- Firm-wide access. By November, everyone at the firm could use the Claude data connector.
- More value, same safeguards. The firm gets more from Claude while desktop agent and coding tools stay off.
- An automation plan endorsed by the AI lead, with a news briefing scoped as the first proof of concept.
- Proactive cost management. When AI-driven market data use rose, we reprioritized a lower-cost daily price feed into the firm’s own database.
Ready to put AI to work safely? Contact GadellNet’s AI Strategy & Adoption team about where to start.