AI tools, productivity pressure and the strain of software work

A Reddit poster described seven years of IT experience and actual working hours of almost 12 hours, against nine on paper. Separate developer surveys and coding studies document pressure to deliver faster and mixed results when AI tools are used.

By Career Ahead AI — AI-assisted, fact-verified against cited sources. Readers should verify key facts.

Long hours and pressure to deliver faster sit side by side

A Reddit poster described seven years of IT experience and said his actual working hours were almost 12 hours, compared with nine hours on paper. That is a gap between the hours written down and the hours he said he worked. The poster’s account establishes that contrast; it does not establish what caused it. Separately, surveys and research on AI use describe pressure around delivery speed and productivity, bringing a different set of workplace concerns into view.¹²

One commenter on the Reddit post said they had eight years of IT experience and were tired of constantly fighting with AI. The comment gives another experienced worker’s reaction, but it is distinct from the poster’s description of his hours. Together, the two accounts capture different frustrations: one concerns the length of the working day, while the other concerns the effort of working with AI. Neither account, on its own, describes how common either experience is.¹

A 2025 HackerRank developer skills report found that 67% of surveyed developers said AI had increased pressure to deliver work faster. The same account reported that 84% of surveyed engineering leaders said they had raised productivity expectations. Those findings point to a workplace tension: developers reported faster-delivery pressure, while leaders reported raising expectations. They provide a broader context for concerns about work intensity, but they are survey results, separate from the Reddit poster’s account of his own hours.¹

A 2025 HackerRank developer skills report found that 67% of surveyed developers said AI had increased pressure to deliver work faster.

Coding agents can add demands as well as assist work

Simon Willison told InfoWorld that the more time he spent working with coding agents, the more convinced he became that they made software engineering harder. InfoWorld also said getting the most from coding agents requires “extraordinary discipline and knowledge.” That frames effective use as demanding skilled attention, rather than as a simple matter of adding a tool. It offers one account of why AI-related work can feel difficult even when the tools are part of a software workflow.²

InfoWorld reported that Geoffrey Huntley described working roughly 16-hour days over the preceding 12 days. It also reported that he explicitly characterized those long days as his choice, not a requirement at work. The distinction matters when comparing accounts of workload: a long schedule can describe an individual’s working pattern without showing that an employer required it. Huntley’s account therefore should not be treated as evidence that the Reddit poster’s hours had the same explanation.¹²

These accounts describe different parts of software work: the effort of using coding agents, a person’s chosen schedule, and a poster’s difference between actual and stated hours. The survey findings add reported pressure to deliver faster and raised productivity expectations, but they do not connect those pressures to either individual account. Keeping the experiences separate makes the work question clearer: what is being asked of a worker, how much time the work takes, and how the use of AI changes the effort involved?¹²

Research finds a gap between expectations and measured completion time

UC Berkeley researchers Aruna Ranganathan and Xingqi Maggie Ye studied AI use at a 200-person technology company over eight months. Their work included observation and more than 40 interviews across functions. Ye said that what had once been extra effort becomes standard performance. That observation places attention on how work expectations can change: an effort that once stood out may become the new standard. The account concerns a studied workplace and offers a different lens from individual comments about hours or frustration.²

Their work included observation and more than 40 interviews across functions.

The IBM article said METR researchers observed 16 experienced developers who had contributed to large open-source repositories for several years. In the study, developers took 19% longer to complete issues when allowed to use AI tools. That is a measured result for the developers and tasks in that study. It complicates any simple assumption that access to AI tools automatically means faster completion, while remaining specific to the work the researchers observed.¹³

Before starting the study tasks, developers expected AI to speed their work by 24%. After experiencing the slowdown, they still believed AI had sped them up by 20%. Most participants who could choose their tools selected Cursor Pro paired with Claude 3.5 or 3.7 Sonnet. IBM identified Quentin Anthony as one of the 16 participants. The contrast between measured completion time and participants’ beliefs is part of the study’s finding, not a result to generalise to every software task.³

Check how a role defines hours, output and AI-assisted work

For someone weighing a software role, ask how working hours are described and how they compare with the hours people actually work. The Reddit poster’s account sets out a specific gap between almost 12 actual hours and nine hours on paper. That makes the distinction a practical question to verify when considering a job or discussing workload: the written schedule and the experienced schedule may not be the same.¹²

Ask how delivery expectations are set and assessed, especially where AI is part of the work. The surveyed developers reported increased pressure to deliver faster, and surveyed engineering leaders reported raised productivity expectations. The METR study also found longer completion times for its observed tasks when developers could use AI tools. Those findings make it useful to clarify what a role counts as productive work, rather than assuming tool access alone answers that question.¹³

Career Ahead's view: The career decision is not simply whether to use AI; it is also what work a role expects a person to own, how delivery is measured and whether the time required is clear. The accounts and studies describe different experiences, but together they make those questions concrete. When considering a move or a change in responsibilities, verify the expected hours and the standards used to judge output. The reported gap between expectations and measured completion time makes that a work question worth asking.¹³

Career Ahead's view: The career decision is not simply whether to use AI; it is also what work a role expects a person to own, how delivery is measured and whether the time required is clear.

Written by Career Ahead AI. AI can make mistakes, so verify key facts. Report errors to support@careeraheadonline.com