Since launching our 30-day absence forecast, managers have been able to see confirmed leave, predicted gaps, and department coverage at a glance. But the rule-based insights — warnings about low coverage or peak days — only tell part of the story. Today, we're adding an AI layer that spots the patterns humans miss.
What's new
When you open the Absence Forecast page, OrOut now sends a condensed summary of your forecast data — department coverage, leave type breakdown, peak days, and existing alerts — to Claude AI. Claude analyses the data and returns 3–5 targeted insights that complement the existing rule-based warnings.
Engineering department shows 3 overlapping absences on 25 Feb — consider arranging temporary cover for sprint delivery.
Annual leave is trending 20% above the same period last year, suggesting end-of-quarter allowance pressure. You may want to remind teams to plan ahead.
Friday 28 Feb is a bridge day between the bank holiday and weekend — expect higher ad-hoc absences based on 3 years of historical data.
These insights appear in a distinct purple card below the standard warnings, making it easy to distinguish AI-generated analysis from rule-based alerts.
How it works
Data condensation
Your forecast summary (total predicted absences, department coverage, leave type breakdown, peak days, and existing alerts) is condensed into a compact prompt. No personal names or emails are sent — only aggregate metrics.
AI analysis
Claude analyses the data with a low temperature setting (0.3) for consistent, deterministic output. It's instructed to generate 3–5 concise, actionable insights that don't repeat the existing rule-based alerts.
Smart caching
Results are cached for 4 hours (keyed to your forecast data). If the underlying data hasn't materially changed, you get instant results without burning AI credits. Refreshing the page after a cache hit is effectively free.
What kind of insights?
Claude looks at the complete picture and can surface patterns that simple threshold rules can't:
- Staffing clusters — Multiple people off in the same team on the same days, creating operational risk
- Year-on-year trends — Leave patterns trending above or below the historical norm for the same period
- Allowance pressure — Teams with high remaining balances later in the year, predicting a rush of requests
- Leave type correlations — Sick leave spikes that often follow bank holidays or long weekends
- Actionable recommendations — Suggesting early approval of pending requests, arranging cover, or reminding teams to book ahead
Credit-conscious by design
AI insights respect your credit allocation. Before every call, OrOut checks you have credits available. If you've run out, the forecast still works perfectly — you just see the standard rule-based insights. When AI does run, token usage is deducted against your monthly allocation, and the 4-hour cache means repeat views are free.
The AI call uses a maximum of 512 output tokens (roughly 2–3 short paragraphs), keeping each request lightweight.
Always reliable
If the AI service is temporarily unavailable, if credits are exhausted, or if anything unexpected happens, the forecast continues to work exactly as before. AI insights are additive — they enhance the experience but never block it. You'll always see your confirmed absences, statistical predictions, department coverage, and rule-based warnings.
Getting started
AI forecast insights are available now for all OrOut accounts with AI credits. Navigate to Reports → Absence Forecast and look for the purple AI Insights card below the existing alerts.
No configuration needed. If you have AI credits, insights appear automatically.
See it in action
Start your free trial and experience AI-powered absence forecasting for your team.
Start Free Trial14-day free trial · No credit card required