See how complete your organisation's knowledge is, then let an AI interviewer ask about the gaps and record your answers as organisation facts for review.
ComplyTrain's AI assistants write better documents and answers when they know your organisation: who you are, how you are organised, which standards and regulations apply, and how you like documents written. This knowledge is kept as organisation facts. AI Discovery measures how many facts you have in each area, shows where knowledge is missing, and runs an interview in which an AI assistant asks about the gaps and records your answers as facts.
Product location: /settings/knowledge-base/discovery. Open Settings, then Knowledge Hub, then AI Discovery.
Admins can open AI Discovery and run sessions. An Admin can let other people run it by adding the Run AI Discovery permission (ai-discovery:run) to one of their roles; see roles and permissions. Everyone else does not see AI Discovery in the navigation.
Every fact recorded in a session becomes an organisation fact that all AI assistants use. Give the permission to the people who know the answers, for example the quality manager and the head of operations, and review the result afterwards.
Product location: /settings/knowledge-base/discovery. Open Settings, then Knowledge Hub, then AI Discovery.
Open Settings, then Knowledge Hub > AI Discovery.
AI Discovery works with an active knowledge base. If your organisation has more than one, choose it under Select Knowledge Base; the default is marked (Default) and is selected first. The session you start is attached to the knowledge base you choose.
What you see | What it means |
|---|---|
No Knowledge Base Found | Your organisation has no knowledge base yet. Create Knowledge Base opens the knowledge sources in the organisation knowledge area; see connect knowledge sources. |
No Active Knowledge Bases | Your knowledge bases are still in draft. Activate one, then return. Manage Knowledge Bases takes you there. |
Product location: /settings/knowledge-base/discovery. Open Settings, then Knowledge Hub, then AI Discovery.
Knowledge Completeness shows one percentage for the organisation, a progress bar and a short summary, such as the number of critical gaps and the facts that are past their review date.
Below it, a card for each category shows its percentage, the number of facts recorded and the number of gaps remaining. The figures are for the selected knowledge base. Each category needs a minimum number of facts:
Category | Covers | Facts needed for 100% |
|---|---|---|
Company Profile | Name, size, industry, locations, products and services. | 5 |
Organization | Departments, teams, key roles and reporting lines. | 3 |
Compliance Context | Certifications, regulations, standards and audit schedules. | 3 |
Preferences | Communication style and documentation format. | 2 |
Terminology | Industry terms, acronyms and internal jargon. | No minimum |
A category's percentage is its number of facts divided by its minimum, up to 100%. The overall figure is the average of the four categories that have a minimum; Terminology is always 100% and is not included. Figures of 80% and above are shown in green, 50–79% in amber and below 50% in red.
Select a category card to prepare a session focused on that category.
Product location: /settings/knowledge-base/discovery. Open Settings, then Knowledge Hub, then AI Discovery.
Knowledge Gaps lists each category that has fewer facts than its minimum. Each gap is named after its category (Company Information, Organizational Structure, Compliance & Regulatory Context or Communication Preferences) and shows the reason, such as Only 1 of 3 minimum facts recorded, and a priority.
Priority | When it is given |
|---|---|
Critical | Compliance Context or Company Profile is below 30%. |
High | Compliance Context, Company Profile or Organization is below 60%. |
Medium | Any category below 80%. |
Low | Any other category below its minimum. |
Gaps are listed with the most urgent first. Choose View all … gaps to open the full list of every gap for the selected knowledge base. Each gap there has Ask about this, which takes the gap into a discovery session so that the assistant asks about it.
Product location: /tenant-context/facts. Open Organization Facts and select Stale only.
Every fact has a review date. Facts Need Review lists facts that are past it, with the number of days overdue, and counts them in its badge.
Review these facts on the Organization Facts page of the organisation knowledge area: select Stale only, and confirm, correct or delete each one; see organisation facts. A discovery session can also ask about them.
Product location: /settings/knowledge-base/discovery/start. Choose Start Discovery Session on AI Discovery.
Choose Start Discovery Session, or select a category card to preselect that category.
Under Choose Focus Area, choose General Discovery to cover every category, or one category. Each option shows its percentage and number of gaps.
Check Gaps to Address, which previews up to five of the gaps the session will cover.
Choose Start Discovery Session. Cancel returns to the dashboard.
The session opens in the chat. Each knowledge base has one session in progress at a time. If an active or paused session already exists for the knowledge base, ComplyTrain asks you to resume it or abandon it before you start a new one. Abandoning keeps the facts that session recorded.
Product location: /settings/knowledge-base/discovery/chat/{sessionId} (fallback: /settings/knowledge-base/discovery). Start a session, or choose Continue on an active session under Recent Sessions.
The header of Discovery Session shows the focus (a category or General Discovery), the knowledge base, and the session's figures: Facts recorded, questions Answered, and gaps Resolved by the session.
Read the assistant's question and type your answer in the message box. Answer in full sentences with concrete details, such as names, numbers and dates.
Send the answer. Processing your response... appears while the assistant works.
The assistant confirms what it has understood, records it as one or more facts, and asks the next question. Correct it straight away if it has misunderstood.
Remaining Gaps beside the chat lists the gaps that are still open, with their priority and reason. It updates as facts are recorded; a gap disappears once its category has enough facts. Choose Ask about this on a gap to have the assistant ask specific questions about it; Asking about: … shows while it prepares them. If you close the panel, you can show it again from the chat.
Each recorded fact:
goes to your organisation facts under the category the assistant chose;
has the source AI Inferred and a confidence level set by the assistant, normally 90%, or 100% when you stated the information outright;
gets a review date, normally six months ahead;
counts towards knowledge completeness straight away.
Product location: /settings/knowledge-base/discovery. Open Settings, then Knowledge Hub, then AI Discovery.
Action | Where | Effect |
|---|---|---|
Pause | Chat header, for an active session | Asks for confirmation, then saves the session as Paused and returns to the dashboard. |
Resume | Dashboard, session list or chat, for a paused session | Makes the session active again and opens its chat. |
Continue | Dashboard or session list, for an active session | Opens the chat. |
Complete | Chat header | Asks for confirmation, then marks the session Completed and returns to the dashboard. You can start a new session at any time. |
Abandon | For an active or paused session | Marks the session Abandoned. It cannot be resumed. |
A paused session does not accept answers until you resume it. If you open a paused session from its row, resume it from the chat. Completing or abandoning a session keeps every fact it recorded; to undo a fact, delete or correct it in Organization Facts.
Product location: /settings/knowledge-base/discovery. Open Settings, then Knowledge Hub, then AI Discovery.
Recent Sessions on the dashboard shows the last five sessions of the selected knowledge base. Each shows its focus, its status (Active, Paused, Completed or Abandoned), the facts extracted, the questions answered out of those asked, and when it was last used. Select an active or paused session to open its chat.
Choose View All to open the list of every discovery session, with its status. Resume or open a session from there.
Product location: /tenant-context/facts. Open Organization Facts and set All Sources to AI Inferred.
Discovered facts are ordinary organisation facts. To review them:
Open Organization Facts in the organisation knowledge area.
Set All Sources to AI Inferred.
Read each fact with its Confidence, Source and Review date. Choose Edit Fact to correct the wording, confidence or review date, giving a Reason for Change, or Delete Fact to remove it.
View History shows who changed a fact and why. See organisation facts and fact proposals and conflicts.
Product location: /settings/knowledge-base/discovery. Open Settings, then Knowledge Hub, then AI Discovery.
Your organisation has just started with ComplyTrain and plans to generate its ISO 9001 documents. The dashboard shows Knowledge Completeness at 10%, with critical gaps for Compliance Context and Company Profile.
The quality manager selects the Compliance Context card and starts the session. They answer questions about the ISO 9001 certificate, its scope, the certification body, the next surveillance audit and the regulations that apply.
They choose Complete, start a new session focused on Company Profile, and answer questions about the legal name, locations, number of employees and main products.
They complete that session too, start a General Discovery session and use Ask about this on the remaining Organizational Structure gap to record the departments and who leads them.
In Organization Facts, they filter on AI Inferred, correct one fact that gave the wrong number of sites, and set a three-month review date for the audit schedule.
With completeness above 80%, they fill in the missing values under document properties and then generate the standard documents.