"Hours drive output."

This is a factory measure applied to knowledge work, where it collapses.

 

On an assembly line an hour genuinely produces an hour of output; for a strategist or a developer, an hour of visible presence tells you nothing about what was solved, created, or prevented. Managing by hours rewards the appearance of work — looking busy — over the thing itself.

 

But the fashionable opposite is also a trap: the pure Results-Only Work Environment (ROWE) removes the one boundary that lets people switch off, and since output is effectively infinite (there's always another email, another doc), the absence of a schedule fills with anxiety rather than freedom. ROWE only works with genuinely clear goals, real feedback, and honest accountability about what "done" means; applied loosely it fails often.

 

The deeper lie beneath both is single-number thinking — the belief that human contribution can be captured in one metric at all. Manage purely by hours and you reward looking busy; manage purely by raw output with no humane limits and you reward burnout. Both extremes just extract more from the worker using a different yardstick.

 

AI is the stress test that breaks both models. Under the punch clock, efficiency gets punished: Ivanti's 2025 Technology at Work Report found that 42% of office workers now use generative AI, and about a third of those users (32%) hide it from their managers — because admitting you did a six-hour task in one hour just earns you five more hours of work, so people pocket the saved time and jiggle the mouse to keep their status green. The punch-clock mentality sabotages the very productivity gains the company says it wants.

 

And under the results-only model, AI drives "the speed-up": when the report now takes an hour, the quota simply rises to six reports by Friday, and the tool sold as liberation cranks the treadmill to a sprint.

Stop chasing a single number; manage the whole picture. Five moves:

 

Define outcomes and boundaries together — agree what "done well" looks like and, in the same breath, realistic timeframes and the right to switch off. An expectation without a boundary is just unlimited work wearing a nicer name. If the scope doesn't fit the hours, cut the scope — not the evenings.

 

Protect the team from the always-on trap — make "good enough, on time" genuinely acceptable. Tell people which tasks deserve deep, perfectionist effort and which just need to be functionally complete. Knowing the difference is what prevents burnout on low-value work.

 

Set humane expectations against the average capable worker, not the outlier — when someone uses AI to turn a two-hour task into ten minutes, don't make ten minutes the new baseline. Research on social comparison shows that benchmarking everyone against the fastest outlier demoralises the whole team and breeds resentment.

 

Reward efficiency with trust, not more work — the cardinal rule of the AI era. If skill or a clever tool saves three hours, let people keep some of that time (leave early, learn something, explore a tool). Fill every saved hour with new tasks and people simply hide the gains forever — and the company loses the entire benefit of the technology.

 

Use input data to support, not surveil — understanding workload and capacity is good management; measuring it by tracking mouse movements or idle-status pings is just the factory punch clock rebuilt with a creepier interface. Support looks like "what's blocking you this week, and do we need to move any deadlines?" — focused on friction in the work, not the location of the worker's hands.