What the AI found
“ "Your sleep quality dips by an average of 18% on nights following work sessions extending past 9 PM, irrespective of total screen time, suggesting the timing of work, rather than just device use, is a significant factor."”
Before
Vague stress, unquantified hunches
After
Clear triggers, data-backed adjustments
The same system, three states — real screens, not a screenshot
| Sleep Duration Avg | 6h 45m |
| Deep Sleep Avg | 1h 10m |
| Workouts / Week | 3 |
| Screen Time Daily Avg | 4h 30m |
Prompt
I have two CSVs: one with my daily sleep quality scores from Apple Health over the last 6 weeks, and another with my work calendar events, including start/end times. Please cross-reference sleep quality against work events ending after 9 PM. Focus on identifying any statistical correlation.
I have two CSVs: one with my daily sleep quality scores from Apple Health over the last 6 weeks, and another with my work calendar events, including start/end times. Please cross-reference sleep quality against work events ending after 9 PM. Focus on identifying any statistical correlation.
AI
Analysis complete. Your sleep quality dips by an average of 18% on nights following work sessions extending past 9 PM, irrespective of total screen time, suggesting the timing of work, rather than just device use, is a significant factor. This pattern shows a statistically significant correlation (p<0.01) across the 6-week period.78% (+12% MoM)
Avg Sleep Quality Score
1.2/week (-60% MoM)
Late Work Sessions (>9 PM)
1h 35m (+25m MoM)
Deep Sleep Duration (Avg)
From Hunch to Hard Data: Unpacking Stress Triggers
An EU marketing manager used a simple AI setup to confirm and quantify the subtle link between late work and sleep quality, transforming vague anxieties into actionable insights.
A 38-year-old marketing manager, EU
Tools used
The real tools used here — swap any for your own equivalent. Each links to how we’d set it up.
Starting state
Before anything was set up
Marta felt perpetually on edge. Her work, while rewarding, often bled into her evenings, leaving her with a nagging feeling of insufficient rest. She suspected late-night screen time was the culprit, but without concrete data, it was just a hunch amidst a sea of daily demands. Her sleep tracking apps offered endless metrics, yet no clear answers about *why* some nights were worse than others. The sheer volume of data felt overwhelming, making it difficult to pinpoint any specific, repeatable patterns in her routine that might be contributing to her stress and poor sleep.
| Sleep Duration Avg | 6h 45m |
| Deep Sleep Avg | 1h 10m |
| Workouts / Week | 3 |
| Screen Time Daily Avg | 4h 30m |
Working state
10-Day Challenge, doing its job
Determined to find clarity, Marta embarked on the 10-Day Challenge. On Day 3, she downloaded her sleep and activity data from Apple Health and exported her work calendar to a Google Sheet. She then uploaded these anonymized datasets into a new chat with Gemini, prompting it to cross-reference late-night work events with her sleep quality scores. The AI quickly processed the disparate information, identifying a subtle but significant correlation she had entirely missed — proving that the AI could connect data points she wouldn't have thought of herself, confirming its utility.
Prompt
I have two CSVs: one with my daily sleep quality scores from Apple Health over the last 6 weeks, and another with my work calendar events, including start/end times. Please cross-reference sleep quality against work events ending after 9 PM. Focus on identifying any statistical correlation.
I have two CSVs: one with my daily sleep quality scores from Apple Health over the last 6 weeks, and another with my work calendar events, including start/end times. Please cross-reference sleep quality against work events ending after 9 PM. Focus on identifying any statistical correlation.
AI
Analysis complete. Your sleep quality dips by an average of 18% on nights following work sessions extending past 9 PM, irrespective of total screen time, suggesting the timing of work, rather than just device use, is a significant factor. This pattern shows a statistically significant correlation (p<0.01) across the 6-week period.Use case implemented
The finished system, running on its own
Armed with this novel insight, Marta adjusted her workflow. She began deferring non-urgent tasks past 8 PM, confident that this small change was directly influencing her well-being. Her weekly review dashboard in Google Sheets, now fed by automated reports from Gemini, clearly displayed her improved sleep quality metrics alongside her adjusted late-work frequency. The system, once complex and overwhelming, now provided clear, actionable feedback, allowing her to make informed choices with a calm, data-driven approach rather than relying on guesswork. The AI had shifted her focus from screen time to work timing, a more impactful intervention.
78% (+12% MoM)
Avg Sleep Quality Score
1.2/week (-60% MoM)
Late Work Sessions (>9 PM)
1h 35m (+25m MoM)
Deep Sleep Duration (Avg)
What an outside observer would notice
Improved by 12% MoM
Average Sleep Quality Score
Reduced by 60% MoM
Late Work Sessions (>9 PM)
Reduced by 80%
Time spent analysing personal data
The stack — build it yourself
Reliable, always-on sensor data for sleep, activity, and screen time, easily exportable.
Standard tool for work scheduling, providing precise timestamps for events.
Easy to import CSVs, perform basic data prep, and create simple dashboards.
Excels at pattern recognition across disparate datasets and summarizing findings concisely.
These are the tools used in this story. Any can be swapped for an equivalent you already trust.
Go deeper
Do this yourself
See 10-Day Challenge
This story runs on 10-Day Challenge. The tools and prompts above are the real build — swap any tool for your own equivalent and follow the same steps.