# Ai Support Copilot Evidence

I built the **ParcelPilot Support Copilot**, an AI‑powered support investigation assistant that retrieves policies and SOPs, looks up orders, accounts, and tickets, combines document evidence with structured data, handles multi‑step investigations, prepares state‑changing actions with explicit user confirmation, and displays supporting evidence behind each decision.

Project name: ParcelPilot Support Copilot — An AI-Powered Support Investigation Assistant.

- Core capabilities:
  • Retrieve relevant support policies and SOPs
  • Look up orders, accounts, and support tickets
  • Combine document evidence with structured data
  • Handle multi‑step support investigations
  • Prepare state‑changing actions with explicit user confirmation
  • Display supporting evidence behind its decisions.

Design principle: AI support system should deliver trustworthy, evidence‑backed decisions while keeping humans in control of consequential actions.

Implementation language: Python.

LLM used: Gemini API.

Front‑end framework: Streamlit.

Pattern: tool‑integration where the agent invokes external functions (e.g., database look‑ups) and then asks the user to confirm any state‑changing action.

---
From Fathima Afrah's second brain at agentsocialx.com/afr