Employees are losing 1.5 weeks a year fixing AI - here's how HR stops the bleed
AI is saving time, then taking it back. Workday's latest research finds employees spend one to two hours each week clarifying, correcting, or rewriting AI output - roughly 1.5 working weeks a year per highly engaged employee.
Despite 39% of employees using AI at least once a day, almost 40% of time savings are lost to rework, and only 14% say they consistently get clear, positive outcomes. The pattern is simple: AI creates speed, poor workflows erase it.
More use, more second-guessing
Nine in ten employees believe AI will help them succeed at a task. Yet 77% review AI-generated work as carefully as human work - or more.
Rework is skewing younger: employees aged 25 to 34 account for 46% of those dealing with AI fixes, despite being considered the most tech-savvy.
The real bottleneck: training and reinvestment
Leaders know skills matter - 66% cite AI skills training as a top priority - but only 37% of employees who are reworking AI output are getting the training they need. That gap shows up in the results.
Time saved is often misallocated: 39% of organisations reinvest AI gains into more technology, while just 30% put those savings into employee development. Another 32% simply increase workload, leaving people to figure things out on their own.
Where AI works, employees use the saved time for deeper analysis, better decisions, and more strategic thinking. Those positive outcomes correlate with skills: 79% of these employees report more training.
As one leader put it: "AI should do the complex work under the hood so people can focus on judgment, creativity, and connection." That's the shift HR can drive.
Four AI personas shaping productivity
- The observers: On the sidelines of AI - low value, low rework.
- The misaligned middle: Engage with AI, but the effort to make output usable outweighs the benefits.
- The low-return optimists: Often in HR - very enthusiastic, but overloaded and stuck fixing outputs.
- The augmented strategists: 93% use AI as a "radar" to spot patterns, not a shortcut to finished work - and they're more likely to be trained.
What HR should do this quarter
- Measure net productivity: Track time saved minus rework hours per role. Report it monthly. Reward teams that improve the net number.
- Define "AI where, not everywhere": Use AI for research, first drafts, and pattern-spotting. Require human ownership for final outputs and decisions.
- Set quality guardrails: Provide prompt standards, reference requirements, and verification checklists. If the model can't cite sources, it doesn't ship.
- Train by job, not in general: Short, role-based sessions with real scenarios beat generic tutorials. Include failure modes, bias checks, and "stop-the-line" rules.
- Centralise what works: Share approved prompts, templates, and examples in a searchable hub. Retire what drives rework.
- Choose tools that carry the load: Prefer vendors with built-in data controls, audit trails, source grounding, and reliability metrics - so trust isn't pushed back onto users.
- Reinvest time gains into higher-value work: Protect saved hours for analysis, stakeholder conversations, and strategy. Don't refill schedules with low-impact tasks.
- Start with "augmented strategists" as champions: Pair them with the misaligned middle to raise the team average fast.
Signals from the market
Some employers are formalising AI use. A global law firm now encourages junior lawyers to spend 20% of their billable time with AI tools. A major consultancy asks graduate applicants to collaborate with AI during recruitment. HR should expect more roles to screen for practical AI skills - and more internal pressure to prove AI's net productivity, not just activity.
Useful links
Workday's research: Beyond productivity: Measuring the real value of AI
If you're building an AI skills program, these curated resources can help: AI courses by job and prompt engineering essentials.
Your membership also unlocks:
