Skip to content

Block-Level Insights

Every block in an audit can have an AI-generated insight that interprets the block’s data points and provides actionable guidance. Block insights are the primary unit of AI analysis — they explain what the numbers mean and what to do about them.

Each block insight contains four components:

A 1-2 sentence summary of what the block’s data reveals. The finding focuses on the most significant observation — whether that is a strength, a weakness, or a notable pattern.

Example findings:

  • “Email authentication is partially configured. DKIM is active but SPF records are missing, which may affect deliverability to certain email providers.”
  • “Workflow naming conventions are followed in 92% of active workflows, indicating strong operational discipline.”
  • “Only 15% of deal pipelines have automated stage-change notifications, meaning sales managers lack visibility into deal progression.”

The finding is factual and data-driven. It references the specific data points that led to the conclusion.

2-3 actionable steps that address the finding. Recommendations are specific enough to act on without being prescriptive about implementation details that vary by portal.

Example recommendations:

  • “1. Add SPF records for your sending domain in your DNS provider. 2. Verify the configuration using HubSpot’s email health tool. 3. Monitor deliverability metrics for 2 weeks after the change.”
  • “1. Review the 8% of workflows that do not follow naming conventions and rename them. 2. Document your naming convention in a shared team resource. 3. Add naming convention checks to your workflow creation checklist.”

Each step is concrete and ordered by execution sequence.

The priority level indicates how urgently this block’s findings should be addressed:

PriorityMeaning
HighIssues that affect core portal functionality, data integrity, or compliance. Address these first.
MediumImportant improvements that will meaningfully impact efficiency or quality. Address after high-priority items.
LowOptimizations and refinements. Address when higher-priority items are resolved.

Priority is determined by the AI based on the severity of the finding, the weight of the affected data points, and the block’s score relative to expectations.

The estimated effort to implement the recommendation:

EffortMeaning
HighRequires significant time, technical expertise, or cross-team coordination. Likely multiple days of work.
MediumA few hours of focused work. May require some technical knowledge or access to specific tools.
LowCan be completed in under an hour. Typically involves configuration changes or simple updates.

Effort helps you and your clients plan implementation realistically. A high-priority, low-effort item is a quick win that should be addressed immediately.

The AI receives the following context for each block:

  • The block name and section it belongs to
  • All data points within the block, their raw values, and their scores
  • The threshold definitions for each data point
  • The overall block score

From this context, the AI identifies the most significant patterns and generates the structured insight. The low temperature setting (0.3) ensures consistency — the same data will produce similar insights across regenerations.

In the audit detail view, each block displays its insight below the data point breakdown. The insight panel shows:

  • The finding text
  • The recommendation steps
  • Priority and effort badges
  • The generation status (generating, ready, or failed)

If an insight has not been generated yet, you will see a “Generate” button. If it failed, you will see a “Retry” option.

You can regenerate any individual block insight at any time by clicking Regenerate on that block’s insight panel. This triggers a fresh AI generation using the current block data. Use this when:

  • The original generation failed
  • You have updated the audit data and want the insight to reflect changes
  • You want a different perspective on the same data

Regeneration replaces the previous insight entirely. The old insight is not preserved.

Block insights can be included in client-facing reports. Each insight must be explicitly published before it appears in a shared report. See Publishing for details on controlling which insights are visible to clients.