What counts as knowledge
Knowledge is the material the AI reads to answer. One knowledge base and one tone of voice serve every channel, so teams do not end up maintaining a separate knowledge base for the chatbot, the agents and the email AI.
- Website and help centerYour public pages and knowledge articles, read as they are published.
- Uploaded documentsManuals, general conditions, onboarding material and other files your team uploads.
- Shared Google Drive folderA folder your team already keeps current, connected as a source.
- Live lookupsCustomer, order, booking and invoice data read from the systems you connect, through native integrations or an MCP server.
Knowledge agents in the AI Supervisor Engine retrieve this material for the specialist handling the request. The how it works page puts the principle in one line: ground on your sources and take actions in the tools you already run.
Access levels keep internal material internal
Each source has an access level. Material marked for internal use helps your team in their own tools without being shared with customers, so internal procedures and pricing notes can sit next to public articles without leaking into a customer reply.
Live data through lookups
Facts about a specific customer, order, booking or invoice do not live in documents, so they come through lookups in the systems you connect: helpdesks, CRMs, ERPs, billing, logistics and commerce platforms, through native integrations or an MCP server. The integrations page lists what is available today.
A lookup’s customer input can be limited to verified values, such as a customer signed in to your portal, an email address confirmed with a one time code, or the phone number on WhatsApp. Without a verified value the lookup does not run, and in an anonymous chat the AI answers from public knowledge only.
When the knowledge does not cover it
Give each specialist hand off conditions for the questions your knowledge does not answer, and for data that is missing or outdated, so the conversation goes to a person with the history and a summary attached instead of the AI guessing. On email and WhatsApp you can also hold every reply for human review until you trust the answers.
A review agent checks every outgoing response against your guidelines before it is sent. How the controls fit together is on accuracy and hallucination prevention.
Keeping the knowledge current
Give every source an owner who keeps it current. When a policy changes, Nova, the AI Operator, can read the new document and draft the update to the specialist and the help center article so they match. Its changes stay drafts until someone on your team publishes them, and Nova never writes to customers.
Articles written for experienced agents can be hard for an AI to read correctly, so the work of a knowledge rollout is often organising what sits in manuals and in people’s heads. A pilot in one topic shows which parts matter first.
Your knowledge stays yours
Your customer data is not used to train AI models for other customers, and the Terms of Service give Typewise no open ended right to train general purpose models on identifiable customer data. Model inference runs in the European Union with zero data retention by the providers. Details are in the trust center.
Where to read next
The platform end to end is on how it works. Security, hosting and compliance are in the trust center. Plans and the per resolution rate are on pricing, and the free trial runs on your real tickets for 14 days with no credit card.