AI Features & Page Metering
How AI summarization, Q&A, topic analysis, translations, and batch operations work.
Last updated July 1, 2026
One unified unit: the Page
Every AI action is metered by Pages (Page Equivalents), where 1 Page = 250 words (a standard legal/transcription page). This keeps billing in the language lawyers already use.
How pages are measured
| Action | Measurement |
|---|---|
| Media (video/audio) | Duration via ffprobe; ~150 words/min → 1 minute = 1 Page |
| Text (PDF/Word/exhibit) | Total words ÷ 250 |
| LLM generation (summary, Q&A, translation, chat, deep dive) | (prompt + response words) ÷ 250 |
| Embeddings (re-embed, active learning) | Source words ÷ 250 |
Single-document AI
- Q&A: ask about a document; the AI cites the source passage.
- Summary: short / medium / long, with estimated pages before generating.
- Topic analysis: extracted topics with relevance scores.
- Custom extractions: define fields, get structured JSON.
- Coding suggestions: responsiveness, privilege, issue codes you accept or reject.
- Translation: translate to another language, preserving formatting.
Batch AI
Select multiple documents and run summaries, topic analysis, extractions, coding, or translation in one job. Each document's pages are summed and you see the per-document estimate before confirming.
The Interrogator chat & MCP
The Interrogator chat answers case-context questions and is metered by prompt + response tokens. MCP manipulations (re-embedding, active-learning re-scoring, clip synthesis) are metered by the words embedded or video minutes exported.
Cost transparency
You always see an estimated Page cost before confirming an action, and every transaction is logged in your page ledger.