AI Insights
Generate an AI-written read of a saved report — what's working, which formats and pillars pull their weight, your best and worst posts, and what to try next.
AI Insights is an AI-written read of a saved report. It looks at your posts and numbers for the report’s channel and dates, then explains what’s working, what isn’t, and what to try next.
It sits at the top of Reports, above Overview on Average. It only appears while a saved report is open. See Saved Reports and Notes.
AI Insights is a starting point for spotting patterns. It only sees the numbers and posts in your report. It doesn’t know about your offline events, ad spend or anything else outside Orsa.
Quick start (5 minutes)
- Open Reports and pick a channel and dates.
- Click Saved Reports, then Save Current Configuration, and save it.
- Click the AI Insights bar to open it.
- Click Generate AI Insights. The bar now says Queued.
- Keep the page open or come back later. The section checks every minute, and when it’s ready the badge is gone and the sections appear.
Example: a coffee shop saves Instagram – last 30 days and generates AI Insights. Later, Format ROI shows reels are 30% of posts but 55% of engagement.
Generate AI Insights
- Open a saved report.
- Click the AI Insights bar to expand it. It says No AI Analysis Yet.
- Click Generate AI Insights.
The button says Queueing… for a moment. The bar then shows Queued and Analysis queued for batch processing.
AI Insights covers the report’s saved channel, date range and competitors. A moving range like Last 30 Days always means the last 30 days as of when the analysis runs. If you changed anything on screen, click Update Current Report first. Until you do, Generate stops with a message asking you to. See Keep the saved dates current.
When the report has competitors saved, you get a competitor comparison instead. See AI Insights when you compare competitors.
How long it takes
AI Insights isn’t instant. Requests are picked up in batches every 15 minutes. The AI then works on it in the background, which can take longer at busy times.
While the page is open, the section checks every minute and shows a message when the analysis is ready.
The badge next to AI Insights tells you where it is:
| Badge | What it means | What to do |
|---|---|---|
| Queued | Waiting for the next batch | Check back later |
| Processing | The AI is writing it | Check back later |
| Failed | Something went wrong | Click Regenerate or Retry Analysis |
| Outdated | The report changed after it was made | Click Regenerate |
| (none) | Ready to read | Click the bar to open it |
You can leave the page while it works. The result is saved with the report, so everyone on your brand sees it.
Regenerate, retry and the Outdated badge
Regenerate sits on the right of the AI Insights bar once there’s a result or a failure. Click it to make a fresh analysis. It’s hidden while an analysis is Queued or Processing.
When you regenerate, the old analysis stays visible while the new one is Queued, and is replaced when the new one is ready.
Outdated. After a result is made, someone may change the report’s saved channel, dates or competitors, or a moving range like Last 30 Days may have moved on. The bar then shows Outdated and Report settings changed since this was generated. A yellow note inside says which dates, channel and competitors it was made for. Click Regenerate for an up-to-date read.
Failed. The bar shows Failed and Analysis failed — click regenerate to retry. Inside, the reason is shown, for example when the AI’s reply couldn’t be read. Click Retry Analysis. It goes back in the queue.
Example: a salon changed its saved report from Last Month to Last 90 Days. AI Insights now says Outdated, Generated for Mar 1, 2026 – Mar 31, 2026 · @salonname. They click Regenerate.
What each section means
Every result starts with a yellow reminder that it’s AI-generated. Under it, you’ll see some or all of these sections. A section is left out when there isn’t enough to say.
Summary. A short read of the period, with a label: ↑ Improving, ↓ Declining or → Stable. Under it:
- Key Insights: the points most worth knowing.
- Standout Winners: what clearly worked.
- Areas to Improve: what clearly didn’t.
Growth Quality. Is growth real or down to one or two posts?
- Raw trend: how your engagement rate and number of posts changed.
- Outlier-adjusted: whether a few standout posts carry most of that change.
- A conclusion on what that means for next month.
Format ROI. For each format, such as reels or carousels:
- Post share: its share of your posts.
- Engagement share: its share of your engagement.
- A verdict and a suggested change.
Example: “Carousel — Post share 40%, Engagement share 18%.” Carousels take a lot of your posts but bring less engagement.
Visual Patterns. Traits seen in the images and videos of your best and worst posts:
- What’s working visually
- What’s not working visually
- Creative brief direction: one line a designer or photographer can act on.
Pillar Opportunity. A pillar with high engagement but few posts, with its Avg ER and Posts. With only a few posts, it says more data is needed first.
Top Performers and Bottom Performers. Your best and worst posts, with their format, pillar, audience and date, and a link to the post. Each says why it probably worked or didn’t, and one thing to repeat or stop.
Recommendations.
- Double Down On: what to do more of, why, and how.
- Experiment With: something to test, why it might work, and how you’ll know (marked ✓).
- Reduce or Eliminate: what to cut back, and what to do instead.
AI Insights when you compare competitors
When the saved report has competitors under Compare against, AI Insights says AI-powered competitor comparison and shows different sections.
Everyone, including you, is compared on public numbers only, so it’s fair. Reach isn’t used. See Compare Competitors.
Competitive Position. Whether you’re ahead or behind, with a label: ↑ Gaining, ↓ Losing or → Stable. Under it: Key Insights, Biggest Opportunity and Biggest Threat.
Content Type Comparison, Pillar Comparison and Audience Comparison. One card per format, pillar or audience. Each shows Your rate and your post count, how competitors did, a verdict, a next step, and high confidence or low confidence. Low confidence means there were few posts to go on.
Competitor Playbook. Tactics competitors use again and again, with links to their posts. Each is marked:
- Cross-validated: two or more competitors do it. That makes it a real pattern.
- Single competitor: only one does it. Treat it as a lead, not a trend.
- adopt: yes, adopt: maybe or adopt: no: whether it fits your brand.
Whitespace Opportunities. Gaps where engagement is high but few accounts, including you, post much.
Recommendations. Double Down On (where you already beat competitors), Adopt From Competitors and Differentiate By (what not to copy).
Ideas for using AI Insights
Start your monthly review. Generate AI Insights on the 1st, then check its points against the numbers below it.
Example: a boutique’s Pillar Opportunity says Styling tips has the highest engagement on 3 posts. They plan two more this month and check again next month.
Brief your designer. Paste the Creative brief direction into your next design request.
Example: “Close-up product shots with one line of bold text.” The designer uses it for the next three posts.
Turn tests into tasks. Each Experiment With item has a ✓ line saying how you’ll know if it worked. Copy it into a task with a due date.
Example: “✓ Carousels with a question on the first slide reach 4% engagement over 4 posts.”
Add your view in notes. Write what you agree or disagree with in the report’s notes. The Send Summary email leads with your notes and backs them up with the numbers. See Saved Reports and Notes.
Tips for the best results
- Use 30 days or more. With few posts, the AI marks most points low confidence and won’t make firm calls.
- Link posts to pillars and audiences first. Pillar and audience points only use posts that have them. See Performance Breakdowns.
- Upload a Meta CSV for Instagram. With Reach, the numbers behind the analysis are more accurate. See Upload a Meta CSV Report.
- Update the report before you generate. AI Insights uses the saved settings, not what’s on screen.
- Check the post count behind a claim. A big change on 2 posts means less than on 20.
- Treat it as a second opinion. Check every point against the numbers below it and what you know about your business.
Common questions
Why can’t I generate AI Insights? It needs a saved report. Without one, the AI Insights card asks you to save the view first. Save the current view, or open one from Saved Reports.
I clicked Generate. Why does it still say Queued? Requests are picked up every 15 minutes, then the AI works in the background. The section checks every minute while the page is open.
Why does it say Outdated? The report’s saved channel, dates or competitors changed after the analysis was made. Click Regenerate.
Why does AI Insights cover different dates from the page? It uses the report’s saved settings. If you changed the range on screen, click Update Current Report, then Regenerate. A saved moving range like Last 30 Days is worked out on the day the analysis runs.
Why are some sections missing? The AI leaves out a section when there isn’t enough to say. For example, there’s no Pillar Opportunity if no pillar qualifies.
Where did my old analysis go after I clicked Regenerate? It stays visible while the new one is queued, then the new one replaces it. If the new one fails, the old one is kept.
Can the whole team see it? Yes. AI Insights is saved with the report, and everyone on your brand can open it.
Does it look at my competitors’ private numbers? No. Competitors are compared on public numbers only. See Compare Competitors.
Can I trust every point? Use it as a starting point. It only sees the numbers in your report, and it can be wrong. Check each point against the numbers below it.
Was this helpful? Tell us what worked or what was missing athello@orsa.social.