Business Query Router

Context

Each NeoVista business originally used separate Power Automate flows, Excel files, Outlook rules, permissions, and maintenance processes to handle inquiries from HTML contact forms.

Problem

As the number of businesses grew, the fragmented setup became difficult to troubleshoot, created inconsistent data handling, and increased the risk of unnecessary access to customer and operational information.

What I built

I consolidated intake behind a single endpoint using Microsoft Power Automate, Outlook, Excel and its API, Microsoft Copilot, Microsoft 365 Admin, and a common request schema. The system masks identifying information before AI processing and assigns each request to an individual associate using live workload data.

How it works

1. 01Receive HTML form submissions through a centralized endpoint and normalize each request into a common schema.
2. 02Mask email addresses and other identifying information before sending query-focused details to the AI decision layer.
3. 03Use Microsoft Copilot and predefined routing boundaries to identify the business, classify intent, and estimate urgency.
4. 04Read the current workload log, rank eligible associates by active assignments, and select an appropriate owner.
5. 05Record the inquiry in the correct Excel data store and send a structured Outlook notification with the owner and expected next action.
6. 06Manage identities, permissions, service configuration, and access boundaries centrally through Microsoft 365 Admin.

Outcome

• Improved workflow productivity by approximately 60%.
• Reduced typical response times from 24–48 hours to 12–24 hours.
• Reduced missed and misdirected inquiries.
• Made flows, permissions, integrations, and routing rules easier to maintain.

What I learned

AI routing becomes safer and more useful when it sees only the information needed for the decision, operates within explicit boundaries, and combines semantic classification with deterministic workload and permission rules.