Jarvis AI

Processes incoming SMS messages and routes them to a pre-built FAQ knowledge base, using intent matching or keyword extraction to identify relevant answers and respond via text messaging. The system maintains conversation state across multiple SMS exchanges, allowing multi-turn interactions without requiring users to install apps or visit web interfaces. Built specifically for the SMS protocol constraints (160-character segments, latency tolerance, no rich media by default).

Accepts FAQ content (likely via web UI, CSV, or API) and builds an indexed knowledge base that enables fast retrieval during conversation. The system likely uses keyword extraction, semantic similarity, or simple pattern matching to map incoming queries to stored Q&A pairs. Indexing strategy determines response latency and accuracy — simple keyword matching is fast but brittle, while semantic embeddings are more robust but require embedding model inference.

Maps incoming SMS queries to the most relevant FAQ answer by comparing the user's message against indexed Q&A pairs using a matching algorithm (keyword overlap, fuzzy matching, or semantic similarity). The system returns the best-match answer or escalates to a human agent if confidence is below a threshold. Routing logic determines whether users get helpful answers or frustrating mismatches.

Maintains conversation context across multiple SMS exchanges, tracking user identity, previous messages, and conversation history within a session. The system uses phone number or session ID to link incoming SMS to prior exchanges, enabling follow-up questions and context-aware responses. State is likely stored in a session store (Redis, database) with TTL-based expiration to clean up old conversations.

Abstracts the underlying SMS provider (Twilio, AWS SNS, or native carrier integration) and routes inbound/outbound messages through a unified API. The system handles phone number provisioning, message queuing, delivery confirmation, and retry logic for failed sends. Integration likely uses webhooks for inbound messages and polling or callbacks for delivery status.

Offers a free tier with limited monthly SMS volume (exact limits unknown) and paid tiers that scale with message volume or conversation count. Pricing model likely uses pay-as-you-go or tiered buckets (e.g., $10/month for 100 conversations, $50/month for 1000). Free tier allows testing without credit card, lowering adoption friction for small businesses.

Provides a web UI for non-technical users to create/edit FAQs, view conversation logs, and monitor chatbot performance. Dashboard likely includes CRUD operations for Q&A pairs, conversation history viewer, and basic analytics (message count, response time). Built for simplicity over power — no advanced features like A/B testing or custom workflows.

Routes conversations to human support agents when the chatbot cannot answer a question or confidence is below a threshold. Escalation likely triggers a notification to an available agent and transfers the conversation context (phone number, history, original query). Agent can then respond via SMS or escalate to phone/email. Handoff mechanism determines whether customers get seamless support or frustrating context loss.