AI Agent for CRM Automation

I want to replace the patchwork of manual steps that move a prospect from our forms and inboxes into the CRM with a single, reliable AI-driven workflow. The goal is pure process optimization: an autonomous agent (or small group of agents) that can read incoming lead data, enrich it, score it, trigger personalised replies, assign tasks, and keep every CRM field perfectly up to date—no human copy-pasting, no delays.

Scope of work
• Map the current lead-to-CRM flow, highlight bottlenecks, and propose an automation architecture.
• Build the agent(s) in a clean, documented codebase—Python, FastAPI, LangChain, or comparable tools are fine as long as they’re well-supported and easy to maintain.
• Integrate with our existing CRM (HubSpot/Salesforce or similar) through its public APIs.
• Add logging, error handling, and basic analytics so I can see volumes, success rates, and latency.
• Commit everything to our private Git repo and supply a quick-start README plus unit tests.

Acceptance criteria
1. Processes 500 test leads end-to-end with no data loss.
2. Average response time ≤ 3 s from lead arrival to CRM update.
3. All code is commented, lint-clean, and deploys via a single command.
4. Short Loom (or similar) walkthrough showing the workflow live.

I’d like an MVP inside two weeks, then a short polish phase. When you reply, tell me which libraries or LLM/AI APIs you’d choose and link to any similar automations you’ve shipped.