Project Xena
An AI assistant that understands thousands of internal documents, so anyone can find answers in seconds.
Challenge
Product specs, repair manuals, SOPs and years of technical documents were spread across different folders and systems. When sales needed to answer a customer’s technical question, or an engineer needed to look up past repair records, they often had to dig through several documents — or just go ask a senior colleague.
Approach
We started with the questions that come up most and take longest to answer, proved the value with a small proof of concept, then expanded step by step:
- Building the knowledge base: We organized the scattered documents, split them into sections and indexed them. Every AI answer cites its original source, so it’s easy to verify.
- Access control: Connected to the company’s existing account system, so each person can only search documents they’re authorized to see, and sensitive data stays protected.
- Built into daily work: Beyond the chat interface, we brought the Q&A capability into support tickets and internal systems, so answers show up where they’re needed.
- Ongoing evaluation: A bank of test questions is used to check answer accuracy regularly, and the system is tuned based on user feedback.
Results
New hires no longer have to track people down for answers, sales can respond to technical questions faster, and senior engineers are freed from answering the same questions over and over. The knowledge base keeps growing, with new document types added over time.