Chat context

Understanding chat context helps you ask better questions and interpret answers correctly.

What Chat can see

From Home

When you chat from Home, ContractsAI can query contract data your account is allowed to access, including:

• Contract metadata (title, type, status, dates, counterparties, etc.)
• Extracted fields and analysis outputs
• Clause/content context in your environment

Access is always permission-scoped:

• It cannot access contracts or fields you do not have permission to view
• It does not use other users’ private chat sessions as context

From Contract Detail (Chat Sidebar)

When you chat from a specific contract page, the conversation is grounded to that contract context:

• Contract-specific details are prioritized
• Clause/source context may be included when available
• Responses are focused on the currently opened contract unless you ask broader questions

From other contextual pages

When chat is launched from a page with active selection/filter context (for example views or invoice-linked flows), the assistant can use that page state as guidance for the response scope.

How context is managed in long chats

• Chat sessions can support many back-and-forth turns
• In long threads, older turns may be compressed/summarized to fit model context windows
• Most recent and most relevant messages are prioritized for response quality

This means very old details may need to be restated if a thread becomes long.

Conversation history and retention

• Conversations are saved in Chat History so you can reopen and continue
• Reopened chats preserve prior thread context, subject to context-window limits
• Visibility follows tenant and permission boundaries (users only see chats they are allowed to access)
• Retention duration and storage policy can vary by deployment or organization settings

Practical limits to keep in mind

• Answers depend on available indexed contract data
• Missing fields, restricted tabs, or disabled modules can reduce answer completeness
• Very broad prompts across large datasets may return summarized results first
• Some requests may require follow-up prompts for full detail

Tips for better answers

• Ask with clear scope:
  “Show active MSAs expiring in the next 90 days.”
• Add filters explicitly: status, contract type, counterparty, date range, owner
• For comparisons, define dimensions:
  “Compare payment terms and renewal notice periods across top 20 MSAs.”
• If output seems incomplete, refine with follow-ups:
  “Include only signed contracts”, “add governing law”, “show exceptions only”
• In long threads, restate key constraints to keep responses aligned

Troubleshooting

If Chat returns errors or says it cannot access data:

1. Confirm you have the required permissions for the relevant module/fields
2. Verify contracts are present and analyzed/indexed
3. Check filters/selections that might be narrowing context too much
4. Retry with a narrower prompt, then expand step by step