Loading…
Chat With Your Data: Internal Wiki
Every company documents processes somewhere. The problem is that employees don't know where to look or don't have time to search. Questions get asked in Slack instead of looking them up. HR answers the same policy question for the fifteenth time. New employees ask colleagues instead of reading the onboarding guide.
A custom AI layer indexed over your knowledge base means asking "How do I submit an expense over $500?" in Slack and getting the exact policy link and procedure steps, instead of pinging HR or scrolling through the wiki sidebar.
Works with Notion, Guru, Tettra, Slab, Coda, Bookstack, and custom wiki systems. Includes a gap analysis dashboard that shows what employees are searching for that isn't documented yet.
Tell us about your knowledge base setup.
Six components that make your wiki answerable in plain language — with citations, escalation handling, and gap analysis.
Ingestion pipelines for Notion (via API), Guru, Tettra, Slab, Coda, Bookstack, and custom wiki systems. The pipeline preserves page structure (headings, nested pages, linked databases, and embedded content) so answers can reference the correct section within a document.
Notion databases and linked pages are treated as first-class content. A Notion page that links to a related database entry can surface both the policy document and the associated exceptions table when a question requires both.
The most common deployment: a Slack bot in the channels where employees ask questions. Someone asks "How do I submit an expense over $500?" in Slack and the bot replies with the policy excerpt and a link to the full page. No new tool to learn.
When the system cannot find a reliable answer with high confidence, it says so explicitly rather than guessing. The escalation path is configurable. It can suggest a human contact, link to the relevant wiki section to browse manually, or create a help desk ticket for the unanswered question.
A dashboard shows every question the system was asked, including the ones it couldn't answer well. This is the most underrated feature: it tells you what your employees are searching for that isn't documented yet. Most clients find two to three significant documentation gaps in the first month.
Pages not updated within a configurable window are flagged in answers so employees know whether they're reading a current policy or a document last maintained two years ago. Stale flagging is configurable per wiki section, some content changes rarely, some should be reviewed quarterly.
Each platform has a dedicated ingestion pipeline. Multi-platform unified search is available for teams that use more than one.
Notion
Via Notion API, including databases and linked pages
Guru
Via Guru API with collection-level access control
Tettra
Via Tettra API with team-level permissions
Slab
Via Slab API with topic and post structure
Coda
Via Coda API including doc pages and tables
Bookstack
Via Bookstack API with shelf and book hierarchy
Custom wikis
Custom wiki systems via REST API or database export
From API connection to live Slack bot with gap analysis, the build sequence.
Connect to your wiki platform
We authenticate to your wiki via API and index the content you specify. You choose which spaces, collections, topics, or databases to include. By default, we index procedural content (policies, guides, SOPs) and exclude meeting notes and personal pages.
Structure and chunk content
Pages are chunked by section heading so retrieval is precise. A long onboarding guide becomes multiple chunks, each retrievable independently. The chunk includes the page title, section heading, breadcrumb path, and last-updated date as metadata.
Configure escalation paths
We define what happens when the system can't find a confident answer. Options: "I don't know: here's the closest section", a direct mention to a subject matter expert in Slack, or a Jira/Zendesk ticket creation. The right path depends on your team structure.
Deploy the bot and test
The bot is installed in your Slack or Teams workspace. We run a test week where the team uses it normally and we collect feedback on answers that were wrong, incomplete, or unhelpful. Retrieval is tuned based on this feedback before full rollout.
Launch gap analysis
The gap analysis dashboard is activated. Every unanswered question is logged. You schedule a monthly review of the top unanswered question categories. This becomes your documentation priority backlog.
Ongoing sync
New and updated wiki pages are synced on a scheduled basis. For Notion and Guru, webhook-based near-real-time indexing is available so new SOPs appear in search within minutes of being published.
Teams with maintained procedural documentation that employees need to reference frequently.
Employees ask HR about vacation accrual, expense policy, parental leave, and equipment ordering dozens of times per month. A knowledge base bot answers these questions in Slack automatically: HR spends time on complex cases rather than repeating policy answers.
Engineers ask about deployment procedures, on-call escalation paths, and internal tool setup in Slack constantly. When the answer is in the wiki but nobody remembers which page, a bot that answers in the channel is faster than a wiki search and a three-minute read.
Support agents field questions from customers that require looking up procedures, product specs, or exception policies. A bot that searches the knowledge base and returns the relevant procedure excerpt (with a link for the full context) cuts average handle time.
We’d rather decline than take a project that won’t deliver value.
Organisations whose wiki content is outdated and unreliable
If your wiki has policies from three years ago that were never updated, procedures that don't match how the team actually works, and pages that contradict each other, adding AI search over that content will return unreliable answers. The content problem must be fixed first. We can help you identify the most critical pages to update before building the search layer.
Teams whose "wiki" is primarily Slack messages and email threads
If institutional knowledge lives primarily in Slack channels and email threads rather than documented pages, a wiki search layer is the wrong tool. The right first step is getting that knowledge into a structured wiki. We can help scope a documentation sprint before building the search layer.
Tell us which wiki platform you use, the approximate number of pages, and the types of questions employees most frequently need to look up. We'll reply within one business day.