Routine Work Analysis
Overview
Your engineering team is not shipping slowly because they lack talent. They are shipping slowly because 84% of their time is consumed by maintenance, bugs, technical debt, and operational toil. The 2025-2026 data is unambiguous: routine work is the single largest tax on engineering capacity, and most organizations massively underestimate its scope.
Key Findings
The Routine Work Breakdown
Chainguard’s 2026 Engineering Reality Report (1,200 engineers and tech leaders across US, UK, Germany, and France) found the time split is far worse than most leaders assume:
| Routine Work Category | % of Developer Time | Source |
|---|---|---|
| Maintenance & technical debt (total) | 84% | Chainguard 2026 |
| Build, deploy & CI/CD toil | 23% | Harness 2025 |
| Managing technical debt | 24% (avg. of workweek) | Sonar 2026 |
| Firefighting & system disruptions | 33% | New Relic 2026 |
| Security tasks & vulnerability remediation | 5% | Harness 2025 |
| New features & writing code | 16% | Chainguard 2026, IDC 2024 |
Two independent sources - Chainguard and IDC - arrived at the same number: developers spend just 16% of their time on application development and new code.
The Maintenance Tax in Hours and Dollars
- 58% of engineering leaders report their developers lose 5+ hours per week to unproductive, automatable work (Cortex 2024)
- 54% of organizations estimate the loss at 5-15 hours per developer per week (Cortex 2024)
- 23% of developer time goes to manual build and deployment toil alone - for a 1,000-developer org, that is 460,000 hours/year or roughly $23M in lost capacity (Harness 2025)
- Technical debt costs an estimated $85 billion per year in global developer opportunity costs (Stripe Developer Coefficient, referenced in 2025 analyses)
- 20-40% of an organization’s entire technology estate value is consumed by technical debt (McKinsey)
Where Routine Time Actually Goes
Cortex’s 2024 survey of 50 engineering leaders at 500+ employee companies identified the top productivity leaks:
- 26% of leaders cite maintenance, bug fixes, and KTLO as their single biggest productivity drain
- 26% cite waiting on approvals and handoffs
- 26% cite gathering project context
- 40% of developers say finding context is their most common daily friction point
The Burnout and Retention Cost
Routine work does not just slow teams down - it drives people out:
- 72% of engineers report struggling to find time for new features due to mounting maintenance demands (Chainguard 2026)
- 52% of developers cite burnout from repetitive tasks as the reason peers leave (Harness 2024)
- 51% of engineers have left or considered leaving a job due to high technical debt volumes (Stepsize)
- 35% of engineers name burnout and excessive workload as their top developer experience issue (Chainguard 2026)
- Only 1 in 3 developers say they spend most of their time on energizing work (Chainguard 2026)
- 93% say feature work is the most rewarding part of their job, yet they get to do it 16% of the time (Chainguard 2026)
Can Automation Actually Help?
The data on AI-assisted maintenance reduction is emerging but promising:
- 89% of organizations using AI tools save 3+ hours per week per engineer; 28% save 5-6 hours (Chainguard 2026)
- 75% of developers agree AI reduces toil time (Sonar 2026)
- 64% of organizations have fully or mostly automated maintenance tasks - and these teams report 94% energizing work ratios (Chainguard 2026)
- TELUS reported saving 500,000 hours across their engineering org using AI agents, averaging 40 minutes saved per AI interaction (Anthropic 2026 Agentic Coding Trends)
- Teams with high automation maturity ship 30% faster (Anthropic 2026)
However, automation is not a silver bullet. Sonar’s 2026 survey found actual toil stays at 23-25% of the workweek even with AI tools, partly because 21-28% of developers now spend time correcting AI-generated code. The shift is real, but incomplete.
What This Means for Your Team
- Measure your actual maintenance ratio. If you have not quantified it, assume it is worse than you think. Chainguard’s 84% maintenance figure shocked even the researchers. Track KTLO hours explicitly in sprint planning.
- Attack build and deploy toil first. At 23% of developer time (Harness 2025), CI/CD and deployment overhead is the single largest automatable category. Every hour saved here directly converts to feature time.
- Set a hard cap on unplanned work. Elite teams on LinearB’s benchmarks keep unplanned work 21% lower than average. Without an explicit budget, maintenance will silently consume your sprint capacity.
- Treat routine work as a retention problem, not just a velocity problem. With 52% of developers citing toil-driven burnout as why colleagues leave (Harness 2024), the cost of inaction is not just slower shipping - it is attrition at $150K-250K per replacement hire.
- Invest in context automation. When 40% of developers say finding context is their top daily friction, chat-native AI agents that can answer “what changed in this service last sprint?” or “what does this error mean?” eliminate a massive class of interruptions.
Sources
- Chainguard 2026 Engineering Reality Report (1,200 respondents, Aug 2025)
- IDC “How Do Software Developers Spend Their Time?” 2024
- Cortex State of Developer Productivity 2024 (50 engineering leaders, 500+ employee companies)
- Harness State of Software Engineering Excellence 2025
- Sonar State of Code Developer Survey 2026 (1,149 developers, Oct 2025)
- New Relic 2026 Data Report
- Harness State of Developer Experience 2024
- Stripe Developer Coefficient Report (referenced in 2025 analyses)
- McKinsey “Breaking Technical Debt’s Vicious Cycle” (referenced in 2025)
- Stepsize Technical Debt Survey (referenced in 2025 whitepapers)
- Anthropic 2026 Agentic Coding Trends Report