AI support answers from documentation, cited and checked
Support teams already wrote the answers: in help articles, product manuals, setup guides and policy pages. ChatDocument turns that documentation into answers for agents and customers, each one cited to the page, and says plainly when the documentation does not cover a question so the ticket goes to a person instead of getting a confident guess.
For support leads, agents and documentation owners
- Help center articles, exported or added as web pages by URL
- Product manuals, installation and setup guides
- Release notes and known issues lists
- Warranty, returns and shipping policies
- Internal troubleshooting playbooks for agents only
Questions that arrive every shift
- How do I reset the device to factory settings?
- Which plans include export to CSV?
- What is the return window for an opened item?
- Why does the sync fail with an expired token error?
- Does the latest release support single sign-on?
Two spaces, one for agents and one for customers
Keep an approved customer space with public documentation only, and an agent space that adds internal playbooks. Connect the documentation folder from Google Drive, SharePoint and OneDrive or Dropbox so updates sync on a schedule.
- Agent workspace: agents ask in the web app, get cited answers and paste the reply with the link to the exact page. Saved threads keep tricky cases for the next shift.
- Help center widget: on Business and Enterprise, add the widget to your site with a script tag. It is restricted to one space, so customers only ever get answers from the documentation you approved. Setup details on document chat widget.
- REST API: on Business and Enterprise, call ask from your ticketing or chat tool to suggest a cited answer on each new ticket. API questions count toward the plan. See document question answering api.
- Compare: check old and new versions of a policy before a change goes live.
- Extract to table: list release notes with columns such as Version, Release date, Feature, Affected plans and Known issue, every cell linked to its page.
The widget says when it does not know
If the approved documentation does not answer a question, the widget tells the customer plainly that the documentation does not cover it, so they know to contact your team. It never fills the gap from general knowledge, so it does not invent features, prices or policies. Agents see the closest passages it checked, which shows where the documentation needs a new article.
Plans for agent teams and public widgets
The widget and API come with Business and Enterprise. Smaller teams use the agent workspace on Starter or Team.
- Starter, 49 USD per month or 24 USD per month billed yearly (288 USD a year): 1 seat, 500 questions a month, 60 per day.
- Team, 149 USD per month or 74 USD per month billed yearly (888 USD a year): 5 agent seats, 2,000 pooled questions, 250 per day, roles.
- Business, 499 USD per month or 249 USD per month billed yearly (2,988 USD a year): 15 seats, 7,000 pooled questions, 800 per day, widget, API and audit log.
- Enterprise, 1,490 USD per month or 745 USD per month billed yearly (8,940 USD a year): 50 seats, 20,000 pooled questions, 2,000 per day, SSO (SAML), SLA 99.9 and invoice billing.
Widget and API questions draw from the same pool as agent questions. See chat with pdf pricing.
What the assistant does not do
ChatDocument answers from documentation. It does not look up an individual customer's order, account or billing, it does not issue refunds, and it does not take actions in other systems. Those stay with your agents. For a wider look at website assistants, see ai chatbot for website.
FAQ
Can the widget answer from internal playbooks by mistake?
No. The widget is restricted to the one space you choose. Put only approved public documentation in that space.
What happens when the daily question limit is reached?
Questions stop for the day until the daily limit resets, and the monthly pool still applies. Higher plans have higher daily and monthly limits.
Can we add help articles by URL?
Yes. Web pages can be added by URL alongside PDF, Word and other files.
Is customer conversation data used to train models?
No. Documents, questions and answers are never used to train models.
Let your documentation answer first
Create your account, upload your help articles and ask the questions your agents hear most. No card needed for the demo.