Maahir Sharma logs off his software engineering job at a Big Tech firm in Dublin. Then the real work starts. The 24-year-old spends about 20 hours a week tinkering with AI tools. He built an agent that calls hotels and negotiates room rates. Others follow suit. They squeeze in experiments between dinner and sleep. Or on weekends.
The After-Hours Grind
This pattern repeats across the industry. Engineers, product leads, and designers pour personal time into mastering systems that promise to transform their output. But the hours add up. They blur lines between office demands and home life. And the pressure comes from a simple fear. Fall behind, and your skills turn obsolete.
According to a survey of more than 1,000 U.S. desk workers across six industries by Ernst & Young, 85% are learning AI skills outside work hours. The findings appear in Business Insider. Workers there describe a race they didn’t sign up for. Yet skipping it feels impossible.
Udit Mehrotra heads product at Amazon in Seattle. In his 30s, he dedicates five to seven hours weekly to AI. Last December he built 10 apps in a month using Claude Code. “I’ve come to think of this less like a sprint and more like a marathon,” he told Business Insider. The pace never lets up.
Abhinav Bohra, a 32-year-old senior applied scientist also at Amazon in Seattle, invests eight to 12 hours weekly. He spent $3,000 last year on tools, conferences, and memberships. “Continuous learning has quietly become part of the job, even when it happens outside the job,” Bohra said. The concern isn’t one tool replacing him overnight. “The bigger concern is becoming technically stale in a field where the baseline is constantly moving.”
Tanvi Pisal knows the stakes. The 29-year-old product designer worked at an AI healthcare startup in San Jose. Laid off in October 2025 amid AI-driven changes, she now contracts for Big Tech as a UX designer. She commits 10 to 15 hours a week to AI study and hundreds of dollars on subscriptions like ChatGPT and Claude. “If I don’t spend a few hours over the weekend catching up on updates, experimenting with tools, or reading about what’s new, I start falling behind,” Pisal explained.
Manoj Aggarwal leads engineering at a large software company. He spends a couple of hours weekly and about $60 a month on subscriptions. His employer offers some access. Still, he trains on his own. Maahir Sharma puts it bluntly. “I think experimentation with AI is very important. If you don’t have hands-on experience, it could be difficult to survive in the industry.”
These accounts reveal a quiet shift. Companies tout AI for speeding tasks that once took months down to days. Productivity metrics climb in some areas. Yet the human cost lands after hours. Workers absorb the learning tax themselves.
But does all this effort deliver? A May 2026 survey of 349 technical workers by METR offers perspective. Participants self-reported a median 1.4 to 2 times change in the value of their work due to AI tools around March 2026. They recalled about 1.3 times uplift a year earlier and projected 2.5 times by 2027. Median self-reported speed gains hit 3 times. The METR report notes these figures come with caveats. Respondents struggle with counterfactuals. Surveys often overestimate gains compared to controlled experiments. METR staff themselves reported the lowest improvements. Selection bias and optimism likely inflate numbers.
Other data paints a mixed picture. ActivTrak’s 2026 State of the Workplace report, drawn from 443 million hours of activity, found 80% of employees now use AI tools at work. That’s up 52% from two years prior. Average time in those tools rose eightfold. Companies average seven or more AI tools now, up from two in 2023. Yet only 3% of users spend the 7-10% of work hours in AI that correlates with peak productivity. The rest see uneven results. (ActivTrak)
DHR Global’s Workforce Trends Report 2026 adds another layer. Nearly 39% of employees report noticeable productivity gains from AI over the past year. Adoption runs strongest in Asia and Europe. North America lags somewhat. Still, leaders often fail to clarify what these gains mean for roles and careers. The disconnect grows. (DHR Global)
Epoch AI’s analysis from April 2026 found half of employed Americans who used AI in the past week apply it at least as much for work as personal tasks. AI has moved from novelty to workplace staple. Yet many still treat it primarily as a personal aid.
The gap between reported speed and actual output shows up clearly in software. A MIT-linked study tracked developers before and after AI adoption. Code output and files edited surged nearly 300%. That dropped to 150% for pieces submitted for review. Full software releases rose only about 30%. Significant. But far from the hype. One X post summarizing the Financial Times coverage noted how perceptions outrun delivered value. (Post by @trevornoren, June 2026)
Faros AI examined data from 1,255 teams and over 10,000 engineers. Individual task completion rose 21%. Commit volume nearly doubled. Review wait times jumped 91%. Time saved individually gets consumed at the next bottleneck. Team delivery barely budges. The missing hours turn into after-work fixes and lingering fatigue.
Harvard Business Review addressed the layoff wave directly in January 2026. Companies cut roles citing AI’s potential, not its current performance. CEOs from Amazon, Salesforce, Ford, and JPMorgan Chase declared many white-collar jobs would disappear. Evidence links generative AI to recent tech layoffs and slowed hiring in marketing and sales. Overall U.S. unemployment stayed low around 4%. The cuts targeted potential displacement more than proven replacement. (Harvard Business Review)
Challenger, Gray & Christmas tracked nearly 50,000 AI-related job cuts announced in early 2026. That represented 17% of total announced reductions. Intuit slashed 3,000 positions, or 17% of staff, to focus on AI. Meta and Cisco followed with thousands more. LinkedIn data showed AI-related postings up 340% since 2024 while traditional software engineering roles fell 15%. Junior and mid-level positions shrank fastest. Seniors with AI expertise stayed in demand.
Amazon offers training, an internal AI hub, and encouragement to experiment on the job. Other firms provide limited support. Many workers buy their own subscriptions and burn personal time anyway. Bohra called it a learning tax. It extracts from evenings, weekends, and mental reserves.
Productivity reports highlight acceleration. ActivTrak found AI increases the pace and density of work. Employees handle more in the same day. But burnout signals rise. One analysis tied AI tools to intensified workloads rather than lighter ones. Developers report mental drain after constant context-switching between human judgment and AI suggestions.
Some engineers note AI acts like an eager junior teammate. Fast but needy. It demands supervision. Experienced hands guide it best. Beginners risk dependency that slows skill growth. The tools boost output on rote tasks. Complex projects still require deep context that AI often lacks.
So workers keep logging those extra hours. Sharma experiments relentlessly. Pisal catches up on weekends to avoid falling behind. Mehrotra treats mastery as a long race. They sense the baseline keeps shifting. Hiring data confirms it. AI fluency commands premiums. Traditional roles contract.
Surveys project bigger gains ahead. METR participants expect value uplift to reach 2.5 times by 2027. Whether that materializes at team or company scale remains open. Bottlenecks in review, integration, and strategy could blunt individual speedups. Organizational inertia often lags the technology.
One thing looks clear. The burden of staying current falls heavily on individuals. Companies reap productivity where it appears. Workers pay in time, money, and recovery. The marathon continues. No finish line in sight.
And the stories multiply. Contractors rebuild careers after AI-linked layoffs. Scientists balance conference budgets with family schedules. Product leaders ship personal apps to stay sharp. Each adds evidence to a trend that defies simple metrics. Output rises in spots. The personal ledger shows different numbers. Hours spent. Sleep lost. Lines between professional drive and personal sacrifice fade.
Recent reporting echoes the tension. A June 2026 CBS News piece detailed how AI job cuts rise but hiring for entry roles weakens even more. The quiet erosion of traditional pathways compounds the pressure on those who remain. They must learn faster, bill more hours, and prove value against systems that improve weekly. (CBS News)
WebProNews is an iEntry Publication