Developers spend 24% of their week on toil, even those who use AI several times a day
Toil is the work that slows developers down: tech debt, debugging poorly documented code, hunting for answers, fixing AI output. Sonar 2026, 1,149 developers, October 2025.
Where developers who use AI several times a day still lose time, by the share who named each a top source of toil (they could pick several, Sonar 2026):
- Managing technical debt: 44%. Every refactor means working out again why the code is the way it is.
- Debugging legacy or poorly documented code: 30%. Nobody remembers why the system works this way.
- Correcting or rewriting AI-generated code: 25%. The AI only saw the task, so the code clashes with the rest of the system.
- Finding information or understanding existing systems: 21%. The answer sits in someone’s head or an old thread.
Why the time is lost
Agents miss the big picture.
- 65% of developers who use AI for refactoring say it misses relevant context. Qodo 2025
- 66% are frustrated by AI answers that are almost right, the most common complaint; 45% say debugging AI code takes longer. Stack Overflow 2025
- 53% worry that leaning on AI will erode the team’s grasp of its own codebase. Sonar 2026
People work from different pictures.
- 81% of software and IT workers say goals and priorities across teams are unclear or conflicting. Atlassian 2026
- Knowledge workers spend a quarter of their week searching for information; 56% say they often have to ask someone or book a meeting to get it. Atlassian 2025
- Only 20% are confident their team has a fast way to tell other teams about decisions that affect them. Atlassian 2025
How we calculate it
Engineering payroll only, tunable per customer:
- Lost today: payroll x 24%.
- Returned with a company brain: 45% of that toil by default (30 to 60% on the pricing page), a multiplier of 0.108 on payroll, about 4.3 hours a week per engineer. It helps most with hunting for answers, poorly documented code and AI rework, and least with tech debt.
Example: a 200-engineer organization
Assuming 200 engineers with an average salary of $160,000 (annual payroll: $32,000,000):
- Lost today
$32,000,000 x 24% = $7,680,000/year of engineering time on toil - Returned with a company brain
$7,680,000 x 45% = $3,456,000/year, about 4.3 hours a week per engineer
Sources
- Sonar, 2026 State of Code Developer Survey (2026)When asked about the portion of their work week assigned to tasks associated with toil (e.g. managing technical debt, debugging legacy or poorly documented code), developers reported spending nearly a quarter of their time on toil work. Interestingly, the amount of time spent on toil (an average of 23-25%) stays almost exactly the same for developers who use AI coding tools frequently and for those who use them less often. (Chart: % of work week spent on toil, Average: 24% of work week, n=1,149, split by Use AI multiple times a day and Use AI occasionally / weekly.) Methodology: Fieldwork for the survey ran throughout October 2025. The final sample size for the study included 1,149 responses, distributed globally.Chart, Toil based on frequency of AI use (n=626). Question: Thinking about your current workflows, what are your biggest sources of “toil work” that hinder / sap productivity or increase frustration in your current role? Use AI multiple times per day / Use AI weekly or occasionally: Managing technical debt (e.g., refactoring complex or outdated code) 44% / 34%; Debugging legacy or poorly documented code 30% / 34%; Correcting or rewriting code created by AI coding tools 25% / 15%; Finding information or understanding existing systems 21% / 29%.Table, Top 10 concerns developers have about AI-assisted or generated code (extremely / very concerned): Over-reliance on AI leading to a decline in the team’s understanding of the codebase, 53%
- Qodo, 2025 State of AI Code Quality (2025)65% of developers using AI for refactoring and ~60% for testing, writing, or reviewing say the assistant “misses relevant context.”
- Stack Overflow, 2025 Developer Survey, AI section (2025)The biggest single frustration, cited by 66% of developers, is dealing with “AI solutions that are almost right, but not quite,” which often leads to the second-biggest frustration: “Debugging AI-generated code is more time-consuming” (45%)
- Atlassian, State of Teams 2026, tech leadership brief (2026)81% of IT and software development workers have unclear or conflicting goals and priorities across teams. Methodology: Double-blind survey data was collected in January and February 2026 from 12,035 global knowledge workers and 173 Fortune 1000 executives.
- Atlassian, State of Teams 2025 (2025)25% Executives and teams alike spend a quarter of the workweek searching for information. 56% of workers say they often find that the only way to get the information they need is to ask someone or schedule a meeting. Methodology: Atlassian surveyed 200 Fortune 1000 executives and 12,000 knowledge workers to discover how the world’s most successful teams are adapting to the changing pace of work.Just 20% of knowledge workers feel confident that their team has an effective process for quickly informing other teams of decisions that may impact their work. Methodology: Atlassian surveyed 200 Fortune 1000 executives and 12,000 knowledge workers.