Project description
We run a growing online service and need a support chatbot that answers customer questions using our existing knowledge base. Right now our support team repeats the same answers dozens of times a day, and response time during peak hours has become a real problem for customer satisfaction. We want an assistant that understands a question phrased in natural language, finds the relevant article, and replies with a clear, accurate answer rather than a link the user has to dig through.
The bot must recognise the intent even when people misspell words or describe a problem in their own way. When it cannot answer confidently, it should hand the conversation over to a human operator together with the full context, so nobody has to repeat themselves. Over time we expect the system to learn from real dialogues: which answers helped, which were rejected, and where the knowledge base has gaps.
Please describe your experience with similar projects, the platforms and models you would use, and how you handle the privacy of customer data. We value maintainability, so clear documentation and a simple admin panel for editing answers are important to us.
What we expect from the finished solution:
— natural-language understanding of typed questions;
— answers sourced from our articles with citations;
— smooth handover to a live operator;
— continuous learning from past conversations.
The bot must recognise the intent even when people misspell words or describe a problem in their own way. When it cannot answer confidently, it should hand the conversation over to a human operator together with the full context, so nobody has to repeat themselves. Over time we expect the system to learn from real dialogues: which answers helped, which were rejected, and where the knowledge base has gaps.
Please describe your experience with similar projects, the platforms and models you would use, and how you handle the privacy of customer data. We value maintainability, so clear documentation and a simple admin panel for editing answers are important to us.
What we expect from the finished solution:
— natural-language understanding of typed questions;
— answers sourced from our articles with citations;
— smooth handover to a live operator;
— continuous learning from past conversations.
Software development Cash, Bank transfer
