The Four Things Driving Your Number
Every estimate traces back to research. Here is what the numbers actually mean.
1. The 23-Minute Attention Residue
Gloria Mark's research at UC Irvine found that after every workplace interruption, it takes an average of 23 minutes and 15 seconds to return to the prior cognitive state. This is called attention residue: part of your cognitive attention stays with the interruption even after you have returned to the task.
Source: Mark, G. (2005). The cost of interrupted work: More than just efficiency. UC Irvine. Presented at CSCW 2006.
2. AI Creates Near-Continuous Micro-Interruptions
Unlike email or Slack (which interrupt in discrete bursts), AI coding assistants interrupt within your flow state. For engineers heavily using Copilot or Cursor, AI-related interruptions can occur 30 to 60+ times per day, each requiring evaluation and context-switching.
Source: Based on Gloria Mark's interruption research applied to AI tool workflows. Observed in our 2,147-engineer survey (2026).
3. Cognitive Load Compounds Asymmetrically by Seniority
The Expertise Reversal Effect (Kalyuga et al., 2003) shows that cognitive support optimized for novices actually hinders experts. Senior engineers who have automated baseline coding processes experience more cognitive load from AI tools, not less.
Source: Kalyuga, S. et al. (2003). The Expertise Reversal Effect. Educational Psychologist, 38(1), 23-31.
4. The Turnover Multiplier
AI fatigue is a top-three predictor of engineers considering leaving the profession (44 percent in our survey). The cost of replacing a senior engineer is 1.5 to 2 times their annual salary when you factor in recruiting, onboarding, and lost productivity. Every preventable departure adds to the real cost.
Source: Society for Human Resource Management; our 2,147-engineer survey data.
Why This Number Is Almost Certainly Understated
The calculator gives you the direct cost: hours lost to attention recovery. But there are indirect costs that are harder to quantify but equally real:
58%
of engineers report measurable skill decline from heavy AI use
63%
feel like middlemen reviewing AI-generated code, not writing it
71%
were treating AI like a test to pass, not a tool to use
44%
were considering leaving the profession entirely
Skill atrophy is a compounding cost. If an engineer loses 5 percent of their coding capacity per quarter due to AI dependency, that lost capacity never shows up in sprint velocity — until it does, when a problem requires a skill they no longer have.
The Measurement Problem
This cost is invisible in conventional engineering metrics. Sprint velocity does not measure cognitive quality. Story points do not capture skill erosion. Lines of code measure output, not understanding. The cost of AI fatigue lives precisely in the gap between what your metrics show and what your engineers can actually do.
The most dangerous costs are the ones that do not appear in your dashboards. A team that ships features on schedule but has quietly lost its engineers' craft is not a high-performing team. It is a team running on borrowed time.
The Clearing, 2026 — based on 2,147-engineer survey data
What Managers Actually See
The signals managers notice are real but easy to misread:
- Engineers who used to volunteer solutions now wait for AI confirmation
- Code reviews