The new bar for engineers

The argument in that YouTube segment is not really about coding. It is about management. AI tools can now draft functions, write tests, summarize pull requests, and scaffold whole features. That changes what counts as ordinary output. GitHub has reported that developers using Copilot finished coding tasks 55 percent faster in one study, and later said users with Copilot Chat completed code reviews 15 percent faster while reporting higher confidence in code quality (github.blog; cacm.acm.org). Once that kind of speed becomes normal, the job does not get easier. The baseline moves. That is already visible in the broader workplace. Microsoft’s 2024 Work Trend Index found that 75 percent of global knowledge workers were using generative AI at work, and 78 percent of those users were bringing their own tools rather than waiting for official approval (microsoft.com). A year later, Microsoft’s 2025 report described the pressure this creates in blunt terms: 53 percent of leaders said productivity must increase, while 80 percent of workers said they lacked enough time or energy to do their jobs (microsoft.com). AI did not remove that strain. It gave executives a reason to ask why more had not changed. That is why the real shift lands on engineering managers. If a machine can remove some of the drudge work, then managers are no longer judged just on shipping a roadmap with the same team. They are judged on whether they can redesign the system around the new capacity. MIT Sloan Management Review put the point cleanly in 2025: companies do not get the promised gains by layering AI onto old jobs and old processes. They get them by deconstructing work, redeploying tasks, and rebuilding workflows around what humans and machines each do best (sloanreview.mit.edu). The manager’s job starts to look less like task assignment and more like operating-system design. That sounds abstract until you look at what happens when companies skip that redesign. Google Cloud’s 2024 DORA report, based on responses from more than 39,000 professionals, found that AI had a significant impact on software development but did not treat adoption itself as the win (dora.dev). The 2025 DORA follow-up made the lesson harder to ignore. AI, it said, acts as an amplifier. It magnifies strong organizations and weak ones alike. The biggest returns come not from the tool itself but from the underlying system around it (dora.dev; research.google). More generated code means more review load, more integration pressure, and more chances for bad process to spread faster. That is where quality risk stops being a side issue and becomes management’s core technical problem. GitHub’s own research on Copilot stressed that speed is only part of the picture, because code still has to be readable, maintainable, reusable, and resilient (github.blog). Amazon’s experience shows what happens when leaders treat AI only as a throughput machine. In a 2025 New York Times report, summarized across multiple republications, Amazon engineers said managers had raised output goals, tightened deadlines, and expected smaller teams to produce similar volumes of code by leaning on AI tools (thestar.com.my; simonwillison.net). The work shifted from writing toward reviewing. The pressure shifted from craft toward pace. That is the new bar. Engineers are still expected to build things, but now they also have to validate, steer, and question machine output. Managers are expected to prove that the time saved turns into something better than more tickets closed. They have to show where the freed capacity went. Better architecture. Faster incident response. More ambitious product bets. Cleaner interfaces between teams. If they cannot point to that higher-order work, then AI productivity starts to look like a spreadsheet trick. The concrete detail is that many engineers now spend their days reading code they did not write, and many managers are being asked to explain why that should count as progress.