AI WhatsApp Sales Follow-Ups

Заказчик: AI | Опубликовано: 05.03.2026

I want an offline-running AI that studies each WhatsApp conversation, decides if a follow-up is actually necessary, chooses the best moment to send it, then drafts and fires off a friendly, contextual message—all without keyword tricks. The single most important benefit for me is saving the hours my team now spends chasing silent leads, so the system must be accurate enough that we can trust it to work on its own. How I picture the flow 1. A new or ongoing chat lands in our WhatsApp Business account. 2. Your model (hosted locally, not in the cloud) pulls the full thread, understands intent, sentiment, and stage of the sale from overall context. 3. If follow-up is warranted, it schedules the outreach intelligently—e.g., after an unanswered quote give it six business days, after a brochure maybe a day—while letting me tweak those rules. 4. At the scheduled time it generates a natural, friendly line or two that feels like a real salesperson checking in, then sends it through our existing WhatsApp API integration. 5. Every action is logged so we can review performance and fine-tune. Key requirements • Offline / on-prem model: no external calls once deployed. • Pure contextual reasoning; no hard keyword lists. • Friendly tone in every autogenerated message. • Easy configuration of timing rules, confidence thresholds, and opt-out logic. • Python preferred; integrate with WhatsApp Cloud API or Twilio’s WhatsApp endpoint (whichever you’re more comfortable with). Deliverables • Trained language model or fine-tuned weights. • End-to-end orchestration code with clear installation guide. • Sample configuration file plus a small set of test conversations demonstrating correct follow-up behavior. • Short README explaining how to retrain or adjust parameters. Acceptance criteria The system must correctly suppress a follow-up on at least 90 % of conversations that already received a response, and must draft human-sounding outreach in 95 % of test cases judged by three uninvolved colleagues. If you have experience with local LLM deployment and WhatsApp automation, let’s talk.