Rate responses and revise with AgentCopy Anchor Link

Every completed AI response in Claira now carries a small row of feedback buttons at its bottom-left corner. Use them to tell us when a response was good or off the mark, and, for document scans, to get concrete prompt improvements generated from your own comment.

Where the buttons appearCopy Anchor Link

The buttons appear only once a response has finished generating (never on errors or while Claira is still working), in three places:

• Prompt Lab, under the AI Response for a single-document scan: thumbs up, thumbs down, and Revise with Agent.
• Multi-Code (single document), under the response window for a one-document Multi-Code run: thumbs up, thumbs down, and Revise with Agent. One rating covers the whole response; Revise with Agent works out which field your comment is about (see below).
• Agent chat, under each completed Agent reply: thumbs up and thumbs down.

Hover over any of the icons to see what it does. The tooltip also notes that response and document data will be shared with the Claira team: your rating, the prompt, and the response are recorded so we can review real examples of where Claira helped or fell short.

Thumbs up and thumbs downCopy Anchor Link

Click 👍 if the response did what you wanted, 👎 if it didn't.

• Your selection fills in with a little pop so you know it registered. There's nothing else to do.
• Changed your mind? Click the other thumb to switch, or click the same thumb again to clear your rating.
• Each rating is tied to one specific response. Running a new scan or sending a new chat message starts fresh.

Revise with AgentCopy Anchor Link

Revise with Agent turns a complaint into a fix. Instead of just marking a response down, you describe what should have been different, and Claira drafts the prompt edits that would get you there. Revise with Agent was called Smart Revise before, which is the name the video above shows.

1. Click the wand icon (Revise with Agent) under the response.
2. In the Revise with Agent popup, describe the change you want in a sentence or two, for example, "The summary should always list the key dates in a bullet list."
3. Click Suggest prompt edits. Claira analyzes your comment against the prompt, the response, and the document.
4. Review the suggestion: each card shows what changed and the complete revised prompt.
5. Click Apply to put the revised prompt straight into your prompt editor, or Copy to take it elsewhere. Nothing re-runs automatically: scan again whenever you're ready.

On a Multi-Code run, your comment is automatically attributed to the right place: if it concerns one field ("the date format is wrong"), only that field's prompt is revised; if it concerns the whole output ("answer in French"), the shared general instruction is revised instead.

Every applied revision is kept in Agent ModeCopy Anchor Link

Applying a suggestion no longer changes your prompt silently. Claira records it in an Agent Mode chat: what you asked for, what changed, and the revised prompt itself, and the dialog offers Open Agent chat so you can open it straight away.

One revise session gets one chat, so if you apply suggestions to three Multi-Code fields you get one thread with three entries rather than three separate chats. From that chat you can insert the revised prompt again at any time, and it goes back to the editor it came from: a field's prompt returns to that field, not to the AI Review prompt. If the field has since been renamed or removed, the entry stays readable and only the insert button disappears.

Because the chat holds the wording you actually used and the reason you changed it, it doubles as a record you can point a colleague at later.

The applied prompt is also saved to Prompt History the moment you apply it, with a small wand that reads Suggested by Revise with Agent, so your team can see which wordings the AI proposed. Open that prompt's details and Open Agent chat takes you back to this chat; only you see the link, since the chat is yours.

What gets sharedCopy Anchor Link

When you rate a response or use Revise with Agent, Claira records the feedback together with the prompt, the response, a reference to the document (its ID and title, never the full document text), and your comment. The Claira team reviews this feedback regularly to improve prompts, templates, and the product itself. Data stays in your deployment's region and is removed if the case's data is deleted.