iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
AI can make you faster on a particular task without proving that it will make you better—or worse—at your job over time. A workplace study found a 14% average increase in issues resolved per hour among customer-support agents with an AI assistant, but gains varied by experience. Evidence on lasting skill loss is much less settled. The practical answer is to use AI to accelerate work while keeping responsibility for the question, the standards, and the final judgment.
What workplace studies actually show
Some tasks get faster, but the gains are not universal
In a study of 5,179 customer-support agents, access to a generative-AI assistant increased the number of customer issues resolved per hour by 14% on average. The gains were concentrated among novice and lower-skilled agents; experienced and highly skilled agents saw minimal effects. The result applies to that workplace, tool, and support task—not every occupation or use of AI. See Brynjolfsson, Li, and Raymond’s study, first issued as an NBER working paper in 2023 and published in journal form in 2025.
A broader Microsoft Research review emphasizes that real-world productivity effects depend on roles, tasks, organizations, adoption, and how people use the systems. One strong result cannot establish that everyone will work faster. The review’s measured workplace findings support a conditional conclusion: AI can help with some work, but the effect depends on the work and the worker. Read the report overview or the full Microsoft Research report.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsLower effort on one task is not evidence of lasting learning
In a small Microsoft study, 40 employee volunteers completed a sales-report task with or without Copilot. Participants using Copilot reported lower perceived mental demand: 30 out of 100, compared with 55 out of 100 in the control group. Researchers found no average difference in a subsequent Stroop score. These are short-term results from one task and a small group; they do not establish whether repeated AI use changes independent skills over months or years.
#1 Best Overall
AI also changes where effort goes. A person may spend less time drafting or searching, but more time deciding whether a task suits AI, breaking it into steps, writing prompts, checking the output, and calibrating trust. The Microsoft report describes this as metacognitive work: understanding the goal, choosing a workflow, judging confidence, and adapting when the result is weak. Lower effort in one part of a task does not mean the whole job requires no judgment.
Does using AI make you worse at your job?
The evidence here does not justify a universal yes or no. A 2026 randomized online experiment examined follow-up performance after AI was removed and reported no worse performance than controls in that experiment. Its task and online sample do not settle skill retention across occupations, nor do they estimate the effect of years of workplace use. The NBER paper is evidence about a particular experimental setting, not a general guarantee that skills cannot erode.
It helps to separate three outcomes that are easy to confuse:
- Immediate speed: How quickly you finish a task with assistance.
- Work quality: Whether the result is accurate, complete, and appropriate.
- Independent capability: Whether you can still do the important parts when AI is unavailable or its answer is wrong.
A gain in the first measure does not automatically establish gains in the other two. Likewise, evidence about a single short-term task cannot answer the long-term capability question for every profession.
How to use AI without outsourcing your judgment
Use AI differently depending on whether the goal is immediate completion or building a skill. These practices are cautious ways to keep quality and learning visible; they have not been proven to prevent skill loss.
For work where speed matters most
- Give AI bounded, checkable work, such as organizing notes or producing a first draft, rather than asking it to own an ambiguous decision.
- Set the quality criteria before reviewing its output: what facts must be right, what sources or constraints apply, and what would make the result usable.
- Check consequential claims and revise the result yourself. Treat fluency as presentation, not proof of correctness.
- Notice whether saved time is being spent on better review or simply on accepting more output.
For work where the skill itself matters
- Try an unaided first attempt before asking AI for a solution, then compare approaches. This makes it easier to see what you understand and what the tool supplied.
- Ask for critique, questions, or an explanation of alternatives instead of a finished answer when you are learning.
- Keep occasional unaided practice for skills you need to retain, especially when you may have to perform without assistance.
- Correct errors and explain why they are errors. Reviewing actively makes your own standards part of the process.
Track speed and quality together
Compare similar tasks done with and without AI. Record elapsed time, important errors, rework, and whether you could explain or reproduce the key decisions unaided. A faster first draft that needs extensive correction may not save time overall; a polished result you cannot verify may be a risk rather than a productivity gain.
Keep the comparison local to your own work. Roles, starting experience, task difficulty, and AI use vary too much for a single published percentage to serve as a personal forecast. Microsoft Research discusses explainability, self-evaluation, co-auditing, and support for task decomposition as possible design responses to the judgment AI use requires. These are useful ideas, not established personal interventions that guarantee skill retention.
How common is workplace AI use?
Adoption figures help explain why this question is becoming practical, but they should not be mistaken for current 2026 estimates. In survey data collected in late 2024, nearly 40% of U.S. adults aged 18–64 said they used generative AI. Among employed respondents, 23% had used it for work at least once in the prior week, and 9% used it every work day. These figures come from Bick, Blandin, and Deming’s NBER paper, revised in February 2025; they describe late-2024 survey responses, not today’s prevalence.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
