Document AI (Summary, Chat, Review)

Document AI is signID’s AI-powered document intelligence layer. It runs on any PDF you’ve uploaded (draft or completed envelope) and offers three primary capabilities:

• Summary: a concise overview of what’s in the document
• Grounded chat / Q&A: ask questions about the document, get answers backed by the actual text
• AI risk review: flag potentially concerning clauses (unusual liability, unclear obligations, etc.)

Document AI is layered on top of the signing pipeline. It never modifies the document, it only reads and analyzes.

Open any uploaded PDF in the Desk and click Summary. signID processes the document (usually within a few seconds for typical contracts) and returns a plain-language summary covering:

• Document type (NDA, service agreement, employment contract, etc.)
• Key parties named
• Core obligations for each party
• Key dates and durations
• Payment terms if applicable
• Any unusual or notable clauses

Summaries are useful for quick review before sending, or for auditing what you’ve already sent.

Grounded chat lets you ask natural-language questions about a specific document. Examples:

• “What’s the notice period for termination?”
• “Does this NDA cover derivative works?”
• “Who’s responsible for indemnification?”

The AI answers using only content from the document (grounded), meaning it does not invent information. If the document doesn’t cover the topic, the AI says so rather than making something up.

AI risk review analyzes the document for clauses that might warrant closer legal scrutiny. Common flags include:

• Unlimited liability or indemnification
• Automatic renewal without notice
• Broad IP assignment
• Non-compete or non-solicitation clauses
• Governing law in an unusual jurisdiction
• Missing termination provisions

Each flag comes with a short explanation of why it’s noteworthy and the specific clause text.

Under the hood, Document AI uses intent orchestration to route your query to the right underlying capability, and a vector retrieval index to find relevant document passages.

What that means practically:

• You don’t need to pick which tool to use, the AI figures out from your question whether you want a summary, a fact lookup, or risk analysis.
• Large documents (100+ pages) work as smoothly as small ones because the retrieval index finds relevant passages instead of stuffing the whole document into the AI.
• Answers cite specific passages so you can verify the source.

signID uses a multi-provider LLM setup: primarily OpenAI (GPT family) and Google Gemini. The right model is chosen based on the task:

• Summarization: fast small model
• Grounded Q&A: mid-tier model with good instruction-following
• Complex risk review: larger, more capable model

This routing is invisible to you as a user, you just ask a question. The multi-provider setup means signID isn’t fully dependent on a single AI vendor’s uptime or pricing.

Some deliberate limits on what Document AI does:

• Never modifies the document. Analysis is read-only. If you want AI to draft a document, use the AI Contract Studio.
• Never sends signer PII to third parties beyond the LLM provider. Document content goes to OpenAI or Gemini for processing; that’s it.
• Not used for training. signID uses provider APIs configured to not retain data for training.
• Not used to auto-populate signer fields. AI never fills in a signer’s data without their explicit action.

Workspace admins can view a conversation log of every Document AI interaction in the workspace. Available from Settings → AI → Conversation log.

Each entry shows:

• User who asked
• Document ID and title
• The question and the AI’s answer
• Timestamp
• Tokens consumed (for billing reference)

Useful for auditing AI usage, reviewing what your team is asking about, and tracking AI spend against your budget.

Every Document AI action consumes tokens and charges your AI Wallet. A free daily allowance covers light usage; heavier use draws from your prepaid balance.

Rough guidance on cost:

• Summarization: cheap (usually a few cents)
• Grounded chat: moderate, per question
• Risk review: highest per document

Set per-request and daily budgets in Settings → AI → Budgets to cap costs. When the wallet runs empty, AI actions are blocked (never silently fail or get charged unexpectedly).