01 the business
Facilities management company
240 staff across 15 sites
How-to guides, SOPs and fixes are spread across SharePoint, Google Drive, Notion and old chat threads. New staff ask the same people the same questions. Old documents contradict new ones.
02 before
The problem
How-to guides, SOPs and fixes are scattered across SharePoint, Google Drive, Notion and old Slack threads. New staff ask the same people the same questions. Older documents contradict newer ones, and nobody can tell which is current.
- People spend time searching or asking around.
- Nobody knows which version is current.
- Experts are interrupted constantly.
03 the build
What the AI does
- Step 1: Indexes the folders, wikis and ticket history you choose, keeping each person's existing permissions.
- Step 2: Answers questions asked in Slack, Microsoft Teams or a browser tab in plain language.
- Step 3: Links every answer to the document or ticket it came from so staff can check it.
- Step 4: Says when it cannot find an answer and suggests who to ask instead of guessing.
- Step 5: Sends a weekly list of unanswered questions and conflicting documents to the content owner.
paused: waiting for a person to approve
How it was put together
- Indexed the chosen folders, wikis and ticket history, keeping existing permissions.
- Answered questions in Teams in plain language.
- Linked every answer to its source.
- Sent owners a weekly list of unanswered questions and conflicting documents.
04 human in the loop
Where a person stays in charge
Each knowledge area has a named owner who reviews the weekly gap report and decides which documents to fix or retire. Folders are only added to the index with that owner's sign-off.
- Each knowledge area has a named owner who approves what is indexed.
- People only see answers from documents they already have access to.
- When unsure, it says so and suggests who to ask.
05 rollout
From kickoff to live in 4 weeks
- Weeks 1–2Choose sources and owners; retire obviously outdated documents.
- Week 3Pilot with two sites.
- Week 4Roll out and start weekly gap reports.
06 after
What changes
Staff find the right procedure without tapping a colleague on the shoulder. Experienced people get fewer repeat questions, and gaps in your documentation become visible instead of hidden.
Assume 240 staff each spending 30 minutes a week searching or asking for information, and the assistant halving that.
240 × 15 minutes = 3,600 minutes, or 60 hours a week returned across the company.
What to measure
- Questions answered with a source link.
- Unanswered questions by topic.
- Documents retired or fixed each month.
- Time new staff take to find answers.
07 evidence
What the research says
Knowledge workers spend roughly a fifth of their working week searching for and gathering information.
McKinsey Global Institute, Capturing business value with social technologies (opens in a new tab) · 2012
62% of people said they spend too much time searching for information, 68% lack enough uninterrupted focus time, and 55% said next steps at the end of meetings are unclear.
Microsoft, 2023 Work Trend Index: Will AI fix work? (opens in a new tab) · 2023
In a survey of 31,000 people in 31 markets, 75% of knowledge workers said they use AI at work, and 78% of AI users bring their own AI tools.
08 lessons
Mistakes to avoid
- Ignoring permissions. Never widen access through search.
- Indexing everything. Old documents create wrong answers.
- No owners. Somebody must fix the gaps.
09 questions
Common questions
- How much time does searching take?
- McKinsey found knowledge workers spend roughly a fifth of their week searching for and gathering information, and 62% of people in Microsoft's 2023 survey said they spend too much time searching.
- Which sources can it use?
- SharePoint, Google Drive, Notion, Confluence, Slack, Teams and Jira.