ClairaClaira Help Desk
AI Review

Single Review

Voir en français

How to scan one document at a time in Claira for testing prompts, spot-checking, and individual document review.

Single Review

Single review is the process of scanning one document at a time through Claira. It is the fastest way to test a prompt, spot-check results, or review an individual document that needs attention. Every scan you run this way gives you immediate feedback, making it the natural starting point before scaling to bulk operations.

When to use single review

  • Testing prompts. Before committing tokens to a bulk scan, run your prompt on a few individual documents to see how the AI responds — single scans are free on every plan.
  • Spot-checking. After a bulk scan, revisit specific documents to verify that results look correct.
  • Reviewing individual documents. When you need the AI's analysis of a particular document outside of a bulk workflow.
  • Iterating on prompt wording. Adjust your prompt, scan the same document again, and compare outputs side by side.

How to run a single scan

Document Viewer
Claira
Coding Panel
  1. Open a document in your Document Viewer pane within Nuix Discover.
  2. In the Claira pane, write a new prompt or select a saved prompt from the Prompt area.
  3. Optional: pick a non-text mode from the Scan as dropdown next to the scan button when the document is image-, audio-, or video-based. Defaults to Text (extracted text). See Media scans for the full list of accepted file types per mode.
  4. Click .
  5. The AI's response appears in the Results area of the Claira pane within a few seconds.
In Prompt Lab, the AI Response area may show formatted text (for example headings, lists, or emphasis). If you Apply that output to a Nuix Discover field, the field type and how Nuix stores the value can strip or simplify formatting, so the text in the field may look plainer than it did in Claira.

That is all it takes. The result is shown immediately in the Claira panel, and if you have a field connected, it is also written to that Nuix Discover field.

Reviewing and editing results

After a single scan, you can review the AI's output directly in the Claira pane before it is finalized. This gives you an opportunity to:

  • Read the full response and confirm it meets your expectations
  • Edit the result if needed before saving it to the connected field
  • Decide whether your prompt needs refinement
Single review never consumes tokens on any plan, regardless of the media type. Whether you're scanning extracted text or running a media scan in image, audio, or video mode, a one-document-at-a-time scan is free. The same applies to single-document Multi-Code runs and to Agent Mode (the lab) — every message in a chat thread is free, and using Multi-Code on a single document is free. Tokens are only consumed by bulk scans (including bulk Multi-Code runs). Multi-Code requires Pro or higher.

In Multi-Code, prompt reuse is available directly in the panel: use View History or Quickstart to insert from history, templates, or the generator—Claira will ask you to pick the shared instruction, an existing field row, or a new row (up to 8), then append or replace if that spot already has text. The Insert into control still sets the default focus for the next action.

Using single review to prepare for bulk scans

The most effective way to prepare for a bulk scan is to build your prompt through repeated single reviews:

  1. Start with a draft prompt. Write your initial instructions and scan a document you already know the answer for.
  2. Compare the AI's result to your expected output. Does it match? Is the format correct? Are the key details captured?
  3. Refine and rescan. Adjust your prompt wording and scan the same document again. Small changes in phrasing often produce noticeably different results.
  4. Test on diverse documents. Once the prompt works on one document, try it on documents that are clearly relevant, clearly not relevant, and a few edge cases.
  5. Move to bulk. When you are confident across your sample, proceed to Bulk Scan.
Pick 10 to 25 documents that represent the variety in your dataset. Include easy cases, hard cases, and documents with minimal text. If your prompt handles all of these well, it will perform well at scale.

Next steps


Need help? Contact us at support@claira.to.

Was this page helpful?