Who Gets the Time Saved by AI?

By Hou Yanjun

Enterprises want employees to use AI for efficiency gains, but employees may hesitate to disclose exactly how much time they've saved. This reluctance is understandable. Consider this: If a task that previously took a full day can now be completed in half a day, what happens when employees voluntarily report this improvement? They might be assigned more valuable work—or see their workload double with no change in compensation or career prospects. Meanwhile, potential job reductions remain unaddressed.

From management's perspective, it's reasonable to expect increased output after investing in tools and deployment. But employees are concerned about how their roles will evolve. If leadership fails to address this disconnect while pushing for AI adoption and knowledge sharing, employees won't feel secure in revealing their productivity improvements.

The question of "who gets the saved time" extends beyond early clock-outs—it involves how enterprises restructure work and recognize employee contributions during this transition.

Consider a hypothetical scenario: A company wants to codify a top performer's expertise into an AI assistant for junior staff. Management sees benefits—fewer repetitive questions for the expert, faster onboarding for newcomers.

But the expert might worry: Their value previously came from being the go-to problem solver. How will their contributions be recognized when judgment methods, case studies, and decision rules are systematized? What becomes their primary responsibility?

Knowledge transfer isn't simple either. While documents can be uploaded, explaining edge cases—when standard approaches don't apply and why—often requires live discussion and real-world validation. Even after deployment, the expert may need to review AI responses, fill gaps, and correct errors.

If original responsibilities remain unchanged while adding these new tasks, with only the project lead credited for "launching a system," knowledge contributors feel overlooked. This undermines incentives for ongoing knowledge sharing.

The solution isn't necessarily complex compensation schemes—start by properly accounting for work. Treat knowledge documentation as formal assignments with allocated time. For ongoing QA, define clear responsibilities reflected in workload assessments and performance metrics. If someone temporarily delivers less on core tasks due to these commitments, don't penalize them using old benchmarks.

Enterprises need to institutionalize individual expertise, but mandating "it's your job" may yield compliant documentation without deeper explanation. While submission completeness is easily verified, sustained willingness to refine tacit knowledge depends on employee engagement.

This disconnect rarely surfaces as outright resistance. Employees may use tools as instructed and submit required materials, but stop proactively identifying gaps or optimizing usage. The system keeps running, while managers wonder why improvement plateaus.

Of course, enterprises investing in AI reasonably expect ROI. Using AI doesn't immunize employees from workload adjustments. The key is transparency: What's added, what's reduced, how contributions are measured post-adjustment.

For example, when a top performer spends less time on repetitive queries, they could handle complex clients or refine the AI assistant. The latter may not directly boost their sales numbers but benefits the team. If evaluations only count individual sales, they'll naturally prioritize personal targets over knowledge curation.

Similarly, as juniors handle more tasks independently via AI, review protocols must adapt. Define which outputs are AI-ready versus requiring human verification—don't delegate tasks without oversight. Otherwise, experts might spend more time fixing errors than they saved.

Workload adjustments shouldn't uniformly apply task-specific speed gains. If drafting drops from two hours to thirty minutes, fact-checking, cross-department coordination, and decision-making likely haven't accelerated equally. Quadrupling targets based on one task's improvement risks quantity over quality.

During pilots, clarify changes upfront: Which repetitive tasks decrease, which new responsibilities emerge, who performs QA, how much time knowledge documentation requires. Concrete plans adjusted through implementation foster better cooperation than vague encouragement with uncertain consequences.

Regarding potential role changes, be honest. While no enterprise can guarantee job security indefinitely, explain pilot objectives, undecided elements, and future communication plans. Promising "no impact" when reductions are planned destroys credibility for future assurances.

Enterprise and employee priorities won't always align. Employees care about income, workload, and career prospects—these aren't mere "resistance to change." Companies seeking ROI must reflect this in concrete work arrangements. Sustained collaboration depends on addressing these real differences—something no training session alone can fix.

How enterprises allocate AI-generated time savings depends on business needs, role requirements, and employee circumstances. But if every efficiency gain just means more tasks—with knowledge sharing and proactive improvements unrecognized—employees will increasingly answer "no" when asked about optimization opportunities.