2026-05-06 –, Inspiration A/B
Your best engineer says they're fine. Their resignation letter arrives two weeks later.
On-call burnout follows predictable patterns. Christina Maslach and the Copenhagen Burnout Inventory mapped them decades ago. Yet most engineering leaders still rely on 1:1 vibes to catch it. Here's the problem: social desirability bias means people tell managers what they want to hear. Fear of appearing weak, culture of heroism, or simply forgetting what "fine" actually feels like.
In this talk, I'll connect burnout research to the signals hiding in your on-call data. After-hours pages correlate to sleep disruption. Consecutive on-call days block recovery. Incident severity drives emotional load. And critically: the same load that's routine for a veteran can break someone six months into their first rotation.
You'll leave with:
• A framework for detecting burnout before it becomes a resignation
• Why observed data beats asking, and when to combine both
• How to have evidence-based 1:1s that get past "I'm fine"
This talk explores why traditional methods of detecting burnout fail, and what actually works.
I'll start with the science: Maslach's burnout model (exhaustion, cynicism, inefficacy) and the Copenhagen Burnout Inventory explain how burnout accumulates over time. On-call engineers are uniquely vulnerable due to unpredictability, sleep disruption, and lack of control.
Then I'll address the elephant in the room: when you ask "how are you doing?" in a 1:1, you almost always get "fine." Social desirability bias, fear of appearing weak, heroism culture, and normalization all conspire to hide the truth until it's too late.
The core of the talk translates burnout research into observable on-call signals: after-hours page frequency, consecutive on-call days, incident severity exposure, and the critical insight that the same load affects people differently based on their baseline.
I'll demonstrate these concepts using On-Call Health (open-source, https://github.com/Rootly-AI-Labs/On-Call-Health), showing how to combine observed data with optional self-reported check-ins to catch problems early.
Attendees leave with a practical framework: what to track, how to have real conversations with data, and how to make rotation decisions they can defend to leadership.
Sylvain Kalache is the Head of Rootly AI Labs. A former Senior SRE at LinkedIn, he co-designed a patented self-healing infrastructure. He later co-founded Holberton, training software engineers worldwide. He also writes about AI, automation, and DevOps for TechCrunch, VentureBeat, and The New Stack.