I want to replace a manual, error-prone data-entry routine with a fully automated Python solution that leans on AI for smart validation and smooth workflow orchestration. The current process involves copying fields from spreadsheets and web forms into our CRM, double-checking them, and then emailing a status report—an hour-long task every single day. Your job is to build a script or small service that watches for new files, cleans and validates the content, writes approved records to the CRM through its REST API, and produces a concise log or Slack message so I always know what happened without opening a spreadsheet. If an entry looks suspicious (e.g., missing phone number, invalid e-mail), the system should flag it and hold it for my review instead of forcing me to sift through everything. Preferred stack and hints • Python 3.x with Pandas or Polars for parsing and cleanup • A lightweight rules engine or LLM call (OpenAI, LangChain, etc.) for “does this look right?” validation • Async requests or an SDK for the CRM API • Optional: Playwright/Selenium only if a browser is unavoidable—API first is better Deliverables 1. Well-commented source code in a private Git repo 2. README with setup instructions and environment variables 3. A short Loom or GIF demo showing the workflow end-to-end 4. One revision round after initial hand-off Acceptance criteria • New files dropped in the watch folder are processed within 2 minutes • 99 % of valid rows reach the CRM with fields mapped correctly • Invalid rows are exported to a separate CSV and listed in the log message • The solution runs on a low-cost Linux VPS (Ubuntu 22.04) I chose workflow optimization and data entry as the primary focus, but a clean modular design would let us extend the same framework later for inventory updates or even lead-generation scraping. Let me know if you’ve done something similar and how quickly you can have a proof of concept ready.