AI product · Prototype / development project
LeadPilot AI
LeadPilot AI is being developed to help teams qualify incoming leads and run follow-ups with LLM-assisted workflows, instead of relying only on manual review of every enquiry.
The problem
Sales and operations teams lose time qualifying repetitive enquiries and chasing follow-ups. Context is often buried in inboxes, so response quality depends on whoever happens to pick up the thread.
The solution
A React and Node.js product that uses LLM APIs, retrieval and agent-style tool calling to classify leads, draft follow-ups and keep a working queue. RAG is used where company or offer context should ground the model rather than leaving it to guess.
Key features
- Lead intake and qualification flows
- LLM-assisted follow-up drafts
- RAG for grounded product context
- Agent-style tool calling for workflow steps
Architecture
React client, Node.js services, LLM APIs for language tasks, RAG over internal knowledge, and AI agents that can call tools as part of qualification and follow-up. Vector storage is used where retrieval is required.
Challenges
- Keeping model output useful and reviewable by a human
- Grounding answers in actual offer and process data
- Designing automation that can stop and hand off to a person
Current status
Prototype / development project. No production usage metrics are published for this work yet.