Skip to main content

Insights Rules

Overview​

The Rules tab is where admins configure Aurora Insights: turn nightly analysis on or off, set the conversation cap per night, toggle the five built-in detection rules, and create custom rules to detect anything specific to your business — competitor mentions, pricing pushback, churn signals, or any pattern you can describe in plain language.

How to Access​

  • Click Insights in the left sidebar.
  • Open the Rules tab.
  • Visible only to admins. Other roles do not see this tab.

Interface Overview​

The Rules page has three stacked cards:

  1. Insights Analysis (top) — the configuration card. Shows whether nightly analysis is currently active and exposes a single button:
    • When enabled: Settings opens the configuration modal.
    • When disabled: Activate opens the same modal.
  2. Built-in Rules — the five detection rules that ship with Aurora. Each has a priority badge, a name, a description, and an on/off toggle.
  3. Custom Rules — your team's own rules. Empty by default. Use Add Custom Rule to create one. Each rule shows a priority badge, a name, the prompt you wrote, edit / delete buttons, and an on/off toggle.

Features & Actions​

Enable or Disable Nightly Analysis​

What it does: Turns the entire Aurora Insights service on or off for your account.

Steps:

  1. On the Insights Analysis card, click Activate (when off) or Settings (when on).
  2. In the modal, toggle Enable Analysis.
  3. Click Activate / Deactivate / Save Changes at the bottom of the modal.

Important notes:

  • Disabling analysis stops new findings from being produced. Existing findings remain visible.
  • After disabling, re-enabling will catch up on missed days — up to seven days of conversation history is reanalyzed on the next run.

Set the Conversation Cap​

What it does: Limits how many conversations the AI analyzes each night.

Steps:

  1. Open the configuration modal (Activate or Settings).
  2. Edit Conversations per run — any value between 10 and 5,000, in steps of 10.
  3. Save.

Important notes:

  • The cap is a safety limit, not a target. If your account had only 80 active conversations yesterday, the AI analyzes 80 — not the cap.
  • The cap drives Aurora Token consumption — see the Aurora Token Usage callout in the modal. Each conversation analyzed consumes one Aurora Token from your account balance.
  • A live estimate of nightly token usage updates as you change the value.

Toggle a Built-in Rule​

What it does: Stops the rule from producing new findings without deleting historical ones.

Steps:

  1. Find the rule in the Built-in Rules card.
  2. Click the toggle on the right.

Important notes:

  • Built-in rules cannot be deleted, only toggled.
  • Disabling a rule does not affect findings that were produced before; they remain on the Dashboard with their existing status.

Add a Custom Rule​

What it does: Tells the AI to look for a pattern you describe in plain language and create findings when it sees that pattern.

Steps:

  1. Click Add Custom Rule in the Custom Rules card.
  2. Fill in the modal:
    • Rule Name — short label that appears on the Dashboard (for example: "Competitor Mention").
    • Priority — Critical, Important, or Info.
    • Rule Description (natural language) — what the AI should look for. Be specific. Example: "Flag conversations where a customer mentions a competitor product or asks to compare prices."
    • Description — short summary of the rule for the dashboard.
  3. Click Save Rule.

Important notes:

  • The rule starts enabled by default and runs on the next nightly analysis.
  • Custom rules are evaluated by the AI on every conversation in the nightly batch (subject to the conversations cap), so each rule you add increases the analytical workload but not the total token cost — tokens are charged per conversation, not per rule.

Edit a Custom Rule​

What it does: Updates the name, priority, prompt, or description of a custom rule.

Steps:

  1. Click the pencil icon on the rule's row.
  2. Make changes in the modal.
  3. Click Save Rule.

Important notes:

  • Edits affect future runs only. Existing findings keep the wording they had when they were created.

Delete a Custom Rule​

What it does: Removes a custom rule permanently.

Steps:

  1. Click the trash icon on the rule's row.
  2. Confirm in the dialog.

Important notes:

  • Findings already produced by the rule remain on the Dashboard.
  • Deletion is permanent — you cannot undo it. To stop a rule without deleting it, use the toggle instead.

Activate Free Analysis (one-time)​

What it does: Schedules a complimentary one-time analysis of up to 1,000 recent conversations with no Aurora Tokens consumed.

Steps:

  1. Open Insights for the first time. If you are an admin and analysis has never run, you'll see the introduction page with a Try It Free card.
  2. Click Activate Free Analysis.
  3. Results are ready the next morning — check the Dashboard tab.

Important notes:

  • The free run is one-time per account. After it completes, future analysis runs use Aurora Tokens normally.
  • Non-admin users see "Ask your account administrator to activate a free analysis for your team" instead of the button.

Built-in Rules Reference​

RuleWhat it detectsDefault priorityAuto-resolves?
Broken PromisesAgents who promised something (a quote, a callback, information) and didn't deliverImportantNo — needs human review
Angry CustomersConversations where the customer expressed significant frustration or angerCriticalNo — needs human review
Missed HandoffsActive conversations with no assigned agent and pending customer messagesImportantYes — closes when an agent is assigned or the customer is answered
Stale ConversationsConversations with unanswered customer messages for more than 24 hoursImportantYes — closes when the customer is answered or the conversation is closed
Unanswered QuestionsSpecific customer questions that the team did not addressImportantNo — needs human review

"Auto-resolves" means the system marks the finding as Auto-Resolved at the start of the next nightly run if the underlying condition is fixed.

Custom Rule Fields​

FieldDescriptionRequiredNotes
Rule NameShort label shown on the DashboardYesKeep it short — appears in lists and filters
PriorityCritical, Important, or InfoYesDrives where the finding appears in the Dashboard's priority groups
Rule Description (natural language)Plain-English description of what to look forYesThis is the prompt sent to the AI. Be specific — see "How to write a good prompt" below
DescriptionShort summary of the rule for the dashboardNoA one-line description of what this rule does
EnabledOn/off toggle—Toggled from the rules list, not the modal

Business Rules​

  • Built-in rules cannot be deleted, only toggled on or off.
  • Disabling a rule (built-in or custom) stops new findings but keeps historical ones on the Dashboard.
  • The conversation cap is a maximum, not a target. Days with fewer active conversations cost fewer tokens.
  • Custom rule prompts are sent to the AI for every conversation in the nightly batch. Vague prompts produce noisy findings; specific prompts produce useful ones.
  • Token consumption is per conversation, not per rule. Adding a third custom rule does not triple your token bill.
  • Configuration changes (enable/disable, cap, rule additions) take effect on the next nightly run, not immediately.

FAQ​

Q: How do I write a good custom-rule prompt? A: Describe a specific behavior you want flagged, in one sentence, in the same language your team writes in. Mention the kind of conversation, what is happening, and why it matters. Example of a good prompt: "Flag conversations where a customer asks about pricing for the Pro plan and the agent does not give a specific number." Example of a bad prompt: "Find sales problems." The first one is concrete enough that the AI can apply it consistently; the second is too vague to produce reliable findings.

Q: Why did my custom rule produce nothing for a week? A: Two common reasons: (1) the prompt is too narrow and there genuinely were no matching conversations; (2) the prompt is too vague and the AI couldn't decide what to flag. Try editing the prompt with a concrete example of what you want caught.

Q: Can a single conversation trigger multiple rules? A: Yes. A conversation that is both stale and contains a broken promise can fairly produce two findings — one of each type. Each rule is evaluated independently.

Q: Why is the conversation cap maxed at 5,000? A: The cap exists to keep nightly token spend predictable. 5,000 conversations is enough to cover even very large accounts. If you hit the cap regularly and need it raised, contact support — but most accounts run with the default of 500.

Q: My custom rule sometimes flags things I didn't intend. How do I fix it? A: Edit the rule and add a clarifying clause that excludes the false positive case. For example, change "Flag price complaints" to "Flag price complaints, except when the customer is asking about a discount they were already told about." You can also dismiss individual findings on the Dashboard to remove them without changing the rule.

Q: Where does my Aurora Token balance come from? A: Aurora Tokens are part of your subscription plan. You can see your balance and top-up options in the Subscription & Billing settings. The Token Usage callout in the configuration modal shows your estimated nightly consumption.