Building a Privacy-First AI CRM Assistant for My Friend's Business

Описание и применение

Система локального AI‑CRM для обработки запросов клиента обеспечивает классификацию, извлечение данных и подготовку ответов, при этом сохраняет конфиденциальность и контролирует расходы.

Задача

Matt’s growing business faced a bottleneck in intake operations, with inconsistent inquiries from email, web forms, and messaging that ranged from sales leads to junk, requiring a system to ingest, classify, extract, and draft responses while protecting sensitive data.

Результат

The system processes inquiries locally, producing structured output and draft responses while ensuring privacy, cost control, and deterministic action execution with human approval.

Рабочий процесс

• Ingest incoming messages from email, web forms, and messaging channels
• Classify each message using a local open‑source LLM into one of four categories
• Extract structured data (names, contact info, business needs) from the message
• Draft a response based on the classification
• Pause workflow and create a pending approval for human review
• Upon approval, update CRM record or send email
• Log audit trail and maintain idempotency

Доказательства

Matt runs a growing business, but his daily intake operation was becoming a bottleneck.

Источник

He needed a system to ingest these inquiries, identify what they are, extract useful structured information, and draft the next response.

Источник

For the Hacktoberfest Weekend Challenge: Build for a Friend, I decided to build him a secure, locally hosted AI CRM Intake Automation System written in Go.

Источник

The LLM is responsible for classification, extraction, and drafting. It does not directly send emails, modify CRM records, or execute arbitrary actions.

Источник

Running an open-weight model locally avoids per-token API costs for every inquiry, including messages that eventually turn out to be spam or incomplete.

Источник