AI-Driven Email Reply Automation

I need an application that reads incoming emails, pulls the requested numbers from my data source, and sends back a clear, professional reply without my involvement. The focus is strictly on automating email responses; later on I may explore data-entry or report-generation features, but the initial milestone is all about accurate, context-aware replies.

Scenario
Stakeholders write asking for KPIs, month-over-month variances, or other numeric summaries stored in a Google Sheet (CSV or a small SQL table is fine, too). The system must:
• detect the intent of the message,
• locate and calculate the right figures,
• draft a human-sounding answer that inserts those numbers, and
• push the reply through Gmail or Outlook via API.

Core requirements
• Natural-language understanding and generation (OpenAI, GPT-4, Claude, or similar).
• Code in Python is preferred; feel free to suggest alternatives if you can justify them.
• Secure handling of credentials and data (OAuth, secrets manager, etc.).
• Simple configuration file so I can map new metrics as they appear.
• Log of every request/response pair for auditing.

Acceptance criteria
A test set of 30 sample queries must be answered with at least 90 % numerical accuracy and a professional tone. I’ll provide the sheet and sample emails once we start.

Deliverables
1. Fully documented source code.
2. Setup instructions and an optional Docker file.
3. Short loom or written walkthrough demonstrating the tool in action.

That’s the whole scope for now—if it runs smoothly, follow-on work is likely.