Project Description
I’m building a new module for our in-house ERP and need a purpose-built AI chatbot that can take repetitive, form-based data entry off the team’s hands. The agent will live directly inside the ERP, not on a public website or app, so every call must respect our existing authentication layer and the data models already in place.
Your task is to design, train and deploy a production-ready chatbot that:
• Accepts natural-language instructions from users inside the ERP UI.
• Extracts the required fields accurately and writes them to the correct tables via our REST/GraphQL endpoints.
• Confirms entries or flags validation issues in real time.
Tech freedom is wide open—as long as the stack is standard and maintainable. Python with FastAPI, Node.js with TypeScript, or a comparable framework is fine, and you can leverage OpenAI, LangChain, Rasa or similar libraries if they speed things up. Just keep in mind we prefer containerised deployment (Docker/K8s) so hand-off is smooth for DevOps.
Acceptance criteria
1. Zero-shot or few-shot model achieves >95 % accuracy on our provided sample forms.
2. Response latency stays under two seconds for a typical request.
3. Clear setup documentation plus a hand-over session so our team can retrain or extend intents later.
If you have prior experience embedding conversational agents inside proprietary ERPs or automating structured data capture, I’d love to see a quick example or demo link when you reply.