AI Productivity at Work: Time Use and Skills Development ======================================================== Evaluating AI productivity requires looking at processing speed as well as how the time saved is used. We explain the difference between automation and augmentation and propose organizational guidelines that account for review workloads and learning opportunities. - Reports of time saved at work through AI do not mean that actual working hours have decreased. - Automation means delegating tasks to machines, while augmentation means expanding human capabilities. - Automating repetitive tasks can lay the groundwork for augmentation through learning new skills. - Evaluations of AI productivity must also include the time colleagues spend on review and rework. - Using the time saved for skills development requires adjusting both work allocation and evaluation criteria. Organizations need criteria for allocating work if time saved with AI is to be spent on developing skills. Simply increasing output may leave no room to learn new tasks. Productivity assessments should also examine how the saved time is used. The survey figures are based on CTA's July 2026 survey and SpaceAdd's October 21 announcement. How is time saved with AI being used? In the surveys reviewed, respondents frequently reported doing more work in the same amount of time. However, increased output and reduced working hours are different outcomes. When reading survey figures, first distinguish between the questions asked and the groups responding. Survey and reference date Confirmed figure Distinction to make when interpreting CTA, July 2026 survey 70% of Korean respondents use AI at work The share of survey respondents who use AI CTA, July 2026 survey More than 8 hours saved per week on average through AI use Savings reported by respondents, not actual times they leave work SpaceAdd, announced October 21, 2026 881 office workers surveyed in total Distinct from the group analyzed for the productivity change question below SpaceAdd, same announcement Analysis of 766 respondents who use AI for at least 1 hour per week The group analyzed for the productivity change question SpaceAdd, same announcement 62.4% do more work in the same amount of time A multiple-response result, not a share of time allocated The CTA figures are confirmed by a Yonhap News Agency report covering the October 14, 2026 announcement. The survey was conducted in July of the same year. The sampling design also needs to be checked before applying these figures directly to all office workers in Korea. Yonhap News Agency's report on the CTA survey In the SpaceAdd survey, 62.4% selected increased workload. This does not mean that 62.4% of the saved time was spent on additional work. It is the result of a question that allowed respondents to select multiple productivity changes. Outsourcing Times' report on the survey results The respondents in the two surveys are not the same group. Therefore, you should not multiply 8.2 hours per week by 62.4% to calculate time spent on additional work. Actual time allocation must be checked against the organization's internal work records. Comparing AI automation and augmentation Automation means having machines perform tasks. Augmentation means expanding human capabilities with AI's help. The two approaches connect when time freed up by automating repetitive tasks is spent on learning. Comparison criterion Use focused on automation Use focused on augmentation Main goal Reduce the time and cost of existing work Expand the range of problems people can solve Report-related work Reduce repetitive drafting and organizing tasks Conduct additional analysis and review the basis for judgments Use of saved time Increase output for the same tasks Learning, experimentation, and exploring customer problems Outcomes to check Completion time, output, errors, and costs New capabilities, quality of judgment, and improvement results Operational concern Review burden caused by increased output Expanding roles without training or feedback Automation itself does not mean job cuts. Staffing is a separate decision made by the organization. Even the same automation tool can produce different results depending on how work is allocated. Augmentation should not be understood only as adding new job duties. Improving the level of judgment in existing work is also a form of augmentation. One example is a report writer using AI to compare alternatives. What has been confirmed about the researcher's proposal? Jan-Emmanuel De Neve's argument is a proposal for an organizational strategy that invests in people's capabilities. In an April 30, 2026 contribution to Oxford University, he explains the risks of an approach centered on job cuts. His main concerns are job insecurity and weakened development paths for entry-level talent. Oxford University's expert commentary A related HBR article was published on April 15, 2026. It discusses the long-term benefits of an augmentation strategy. The article title and public summary are available on HBR. HBR's article on automation and augmentation strategies This should not be read as proof that augmentation is superior in every organization. Public commentary alone cannot be used to calculate revenue growth rates by industry. The operational guidelines below are suggestions for applying this discussion to work. A situational guide: Where should you reallocate time? Decide how to use saved time based on bottlenecks in your current work. If there is a backlog, you may need to do more of the existing work. If the problem is a lack of judgment or skills, you can allocate time to learning and review. The cases below are examples of possible applications, not cases with measured results. Decide how to expand your role based on the purpose of your work. This does not mean immediately taking on the responsibilities of another role. Current situation Example of how to use the time Conditions to check alongside it A designer can only explain to developers how to implement an idea Use AI to build a simple working prototype to communicate the intent Developer review before incorporating it into the actual service A salesperson repeatedly requests revisions to materials Use AI to create drafts of explanatory materials directly Check the accuracy of product information and wording Report writing takes up much of the available time Focus on checking evidence and comparing alternatives after drafting A person responsible for checking sources and calculation results Entry-level employees who handled repetitive tasks have fewer opportunities for hands-on practice Practice finding errors in AI outputs and explaining the basis for judgments Feedback from senior colleagues and opportunities to make corrections Learning opportunities for entry-level employees Entry-level employees need a process for learning the criteria used to review work. Even if basic tasks are assigned to AI, design judgment training separately. Do not assume that learning has taken place simply because employees have been assigned review work. For example, you can ask them to identify sentences in a draft report that lack supporting evidence. Then set aside time for a senior colleague to check the reasons for their judgments. Training results can be assessed through their ability to explain and make corrections. Individual time savings and team time savings A reduction in one person's writing time does not necessarily mean that the team's total working time has decreased. The review burden on colleagues receiving the work must also be counted. This distinction is easy to miss in output-focused evaluations. Even if a report draft is produced quickly, fact-checking may take time. More drafts mean more documents to review. It is therefore more appropriate to define the end of a task as final approval rather than draft generation. Measurement What to record Purpose Writing time Time from preparing materials to completing the draft Check the writer's time savings Review time Time colleagues and managers spend checking the work Check the burden shifted to others Rework time Time spent correcting errors and resubmitting Check the actual cost of completing the work Final quality Omissions, errors, and whether revisions were needed after approval Check whether increased speed harmed quality Use of freed-up time Additional work, learning, experimentation, and breathing room at work Check how saved time was allocated A simple check is to subtract the total time spent after adopting AI from the previous total time spent. Include review and rework in the time spent after adopting AI. This is a management approach proposed in this article, not an officially recognized formula. Comparisons must use the same scope of work and quality standards. It is not appropriate to directly compare the result of producing a shorter, simpler report with the previous work. Record reduced time and reduced output separately. How organizations can turn time into capabilities To secure time for skill development, first adjust the rules for assigning additional work. If you encourage training while continuing to increase workloads, learning may be pushed aside. The following steps are suggestions to adapt to your team's work situation. Identify bottlenecks in current work. Ask team members separately about repetitive tasks and improvement projects they have put off. Set goals for AI use. Specify whether you will evaluate increased output or skill development. Choose a small experimental task. Start with a scope that allows you to check the results. Include learning time in the work schedule. Also decide which existing tasks to adjust when new requests arrive. Connect the people responsible for training and review. Teach the criteria for judging results as well as how to use the tools. Evaluate the experiment's results together. Record time, quality, and newly acquired skills. If you cannot easily adjust your own schedule, first agree on work priorities with your team leader. A brief outline of the improvement project's purpose and the support needed can help. You can propose an approach that does not leave learning solely to employees' personal time after work. Managers should also reflect development goals in evaluation criteria. If only output is rewarded, team members will spend their time increasing output. Include the basis for judgments gained through experiments and reusable ways of working in the evaluation. Common mistakes when evaluating AI productivity Describing productivity with a single number can obscure changes in time and quality. The items below are distinctions to make when interpreting survey results and organizational practices. The same distinctions can be applied to team performance reports. Describing time savings as a reduction in working hours. Check actual work records separately. Reading a multiple-response percentage as a share of time allocated. Check the question asked and the group responding. Immediately treating increased workload as expanded capability. Record specifically what new tasks people can perform. Treating AI-generated work as evidence of acquired expertise. Also check whether people can explain and correct errors. Setting the same proportion of learning time for every employee. The reviewed materials contain no universally recommended proportion. Assuming that adopting tools alone completes an organization's growth strategy. Review work allocation, training, and evaluation criteria together. FAQ Q. How do AI automation and augmentation differ? A. Automation means handing tasks over to machines. Augmentation means expanding people's capabilities with AI's help. When you automate repetitive tasks and then use the time freed up for learning, the two approaches connect. Q. If AI saves 8.2 hours per week, does that also reduce working hours? A. That figure alone cannot tell us whether working hours have decreased. It is the average time saved reported in CTA's July 2026 survey. Actual changes in working hours need to be checked against task allocation and work records. Q. Whose responses make up the 62.4% who say they get more work done? A. This figure appears in news coverage of SpaceAdd's October 21, 2026 announcement. The survey included 881 workers in total. The analysis for this question covered 766 people who use AI for at least 1 hour per week. Respondents could select multiple productivity changes. Q. Can I multiply 8.2 hours by 62.4% to calculate additional work time? A. You should not multiply these two figures to calculate additional work time. They come from different surveys. The 62.4% is not a time allocation percentage, either. Q. Is it wrong to use AI to do more of the same work? A. It can help clear backlogs or meet required work volumes. However, increased output and the acquisition of new capabilities should be distinguished. First, decide what outcomes the organization wants. Q. How much of the time saved should be spent on learning? A. The materials reviewed do not provide a universally recommended percentage. Allocate the time based on your current workload and the capabilities you need. You should also check whether learning time is actually included in the schedule. Q. Does AI augmentation mean taking on an additional job role? A. It does not just mean adding another job role. Analyzing issues more deeply or evaluating alternatives within existing work are also forms of augmentation. When expanding roles, provide training and review support as well. Q. Whose time should be included when measuring AI productivity? A. Include not only the author's work time but also colleagues' review and rework time. The tasks being compared must have the same scope and quality standards. Recording the total time spent through final approval can reveal shifts in the workload. Q. How do new employees learn if AI handles basic tasks? A. You can create assignments that ask them to find errors in AI outputs and explain the reasoning behind their judgments. Provide feedback from senior colleagues as well. Receiving an output and developing judgment should be distinguished. Q. Has choosing augmentation been proven to improve long-term performance? A. The publicly available commentary reviewed proposes an augmentation-focused organizational strategy. It does not guarantee the same outcomes for every organization. The actual effects need to be assessed through changes in the organization's time use, quality, and capabilities. Sources - Yonhap News Agency, CES organizer highlights S. Korean firms' AI innovation ahead of 2027 show, October 14, 2026: https://en.yna.co.kr/view/AEN20261014005300320 - Outsourcing Times, 95% of office workers see improved productivity after using AI, October 22, 2026: https://www.outsourcing.co.kr/news/articleView.html?idxno=204137 - Oxford University, Expert Comment: The firms racing to replace people with AI will be the first to fall, April 30, 2026: https://www.ox.ac.uk/news/2026-04-30-expert-comment-the-firms-racing-to-replace-people-with-ai-will-be-the-first-to-fall - Harvard Business Review, Why Companies That Choose AI Augmentation Over Automation May Win in the Long Run, April 15, 2026: https://hbr.org/2026/04/why-companies-that-choose-ai-augmentation-over-automation-may-win-in-the-long-run Images - A man explains something on a tablet to a woman in a design studio with material samples on the table.: https://injoys.com/rails/active_storage/blobs/proxy/eyJfcmFpbHMiOnsiZGF0YSI6MzUyMTIsInB1ciI6ImJsb2JfaWQifX0=--f088ae35758add8f3ffeace7fc5e6d09cccb4b9e/ai-352d59d4.webp - A middle-aged man marks corrections on a report at a sloped desk, checking it against another document.: https://injoys.com/rails/active_storage/blobs/proxy/eyJfcmFpbHMiOnsiZGF0YSI6MzUyMTMsInB1ciI6ImJsb2JfaWQifX0=--1ae1f89d020df24fc4db58e08152298eac5c3d00/ai-4e5fcfb8.webp --- Category: Knowledge Base Source: https://injoys.com/en/articles/ai-productivity-time-use-and-capability-growth License: cc_by Translation-Status: reviewed