Closing the knowledge gap

Answers no one could find

Internal knowledge was spread across SharePoint, Google Drive, Notion, Office documents and an internal knowledge base — five places that rarely agreed with each other. New starters took months to find their feet.

It hit the front line hardest. The Patient Engagement Team was held up on live enquiries, unable to surface the right policy or service information quickly enough — right when people needed a fast, confident answer.

A private assistant they can trust

We built a private Retrieval-Augmented Generation (RAG) knowledge assistant over a six-month engagement, with the first version in production from week four and an iterative build after that. A core team of six delivered it, with security and change-management specialists brought in as needed.

It sits on a secure data layer that indexes SharePoint, Google Drive, Notion, Office documents and the internal knowledge base, respecting existing permissions so people only ever see what they’re already allowed to. It answers only from those verified internal sources — never the open web.

Queries are also routed by sensitivity: general questions can draw on frontier models for stronger reasoning, while anything sensitive stays on a UK-hosted model that never leaves the organisation’s own environment.

What changed

• Time to find an internal answer fell from an average of 12 minutes to under 30 seconds.
• New-starter induction was cut by three weeks.
• First-contact resolution rose 22%, with fewer enquiries needing escalation.
• Five separate tools became one place to ask.

Time redirected to care, not searching

The real dividend isn’t the efficiency stats on their own — it’s what staff do with the time back. Less of every shift spent hunting for the right policy or procedure means more of it spent with the people who need support. That’s the return that actually matters here.