Build Agentic Company Knowledge AI
Publicada el 2026-08-01
Descripción de la oferta
My goal is to turn our scattered product specs, policy docs, and support notes into a single, searchable “company brain” that an agent-style AI can query and then act on. Once the knowledge layer is in place, I want the same system to handle routine operational work: answering emails, taking or routing customer calls, and even placing simple orders automatically. At the knowledge level the AI must comfortably surface: • Product information • Company policies • Technical support material On the automation side it should: • Draft or send email replies that reflect current policy and stock status • Accept or transfer incoming phone calls with natural voice prompts • Generate and place straightforward orders in our back-office system I imagine a retrieval-augmented approach (vector store + LLM such as GPT-4), wrapped in an orchestration framework like LangChain or similar, with hooks into our email, telephony, and ordering APIs. You are free to propose the exact stack as long as it is well documented and cost-conscious. Deliverables 1. Ingestion pipeline that continuously indexes our documents and updates embeddings 2. Conversational interface (web or Slack) for internal testing that demonstrates accurate answers from the three knowledge areas above 3. Workflow scripts/agents that: • Reply to a sample email thread correctly • Simulate an inbound call flow with voice or text • Create a test order via API and return confirmation 4. Deployment guide and brief training session so my team can maintain and extend the system Acceptance criteria: the demo environment must answer at least 90 % of a supplied test set correctly, complete the three automation flows without human intervention, and log every action for audit purposes. Let me know your proposed architecture, timeline, and any prior work with agentic LLM systems so we can get started quickly.
Skills
Fuente original: freelancer