AI productivity is no longer just a promise from tech companies. The 15% figure comes from the study Generative AI at Work, published as an NBER working paper by Erik Brynjolfsson, Danielle Li and Lindsey R. Raymond. The researchers analysed more than 5,000 customer support agents and found that access to an AI assistant increased productivity by 14/15% on average, measured as issues resolved per hour. Rounded, this is the “about 15%” gain often cited in articles about AI and work. That number explains why companies, managers and workers are paying so much attention to AI in 2026.
But the headline does not tell the whole story. AI does not make every worker 15% better in every job. The biggest gains appear when the task is repetitive enough for AI to support it, but complex enough that human judgment still matters. That is why the real question is not simply whether AI increases productivity. It is where, how and for whom it works best.
Contents
Where the 15% Productivity Number Comes From
The 15% figure comes from a study on generative AI used by customer support agents. The researchers looked at more than 5,000 workers and measured how access to an AI assistant changed performance. The main result was clear: workers using AI resolved more issues per hour than those without it.
This is important because it moves the debate away from pure speculation. For years, people talked about AI as something that might improve work one day. Now there is growing evidence that, in the right context, AI can already make workers faster. Customer support is a perfect test case because it includes recurring questions, standard processes, language work and the need to give useful answers quickly.
The Biggest Gains Go to Less Experienced Workers
The most interesting part of the study is not only the average gain. It is the distribution of that gain. Less experienced and lower-skilled workers improved the most. In practice, AI helped them perform more like experienced colleagues because it gave them instant support, better phrasing, faster access to information and guidance on how to handle unusual cases.
This matters because it changes the way we think about workplace training. AI is not only a tool for top performers. In some cases, it can become a training layer that helps newer workers learn faster. Instead of waiting months to absorb knowledge from colleagues, manuals and trial-and-error, workers can receive support while doing the task itself.
Why AI Makes Some Work Faster
AI increases productivity when it reduces the time spent on low-value friction. That can mean searching for information, rewriting answers, summarising documents, drafting messages, checking tone, generating first ideas or turning raw notes into structured output. These are not always the most glamorous parts of a job, but they consume a huge amount of time.
Gallup’s workplace data points in the same direction. Workers who use AI often say they use it to consolidate information, generate ideas and learn new things. These are exactly the kinds of tasks where AI can reduce mental load and shorten the distance between input and output.
Why the Productivity Boom Is Not Automatic
The 15% number is powerful, but companies should not read it as a universal guarantee. AI does not automatically create productivity just because a company buys a subscription. It needs the right workflow, the right data, the right training and the right rules for human review.
This is where many organisations still struggle. Workers may use AI informally, but managers may not know how to redesign processes around it. Some teams save time on drafts but lose time checking mistakes. Others generate more output but not better outcomes. That is why the productivity story is real, but uneven.
Who Benefits Most From AI at Work?
The workers who benefit most are usually those whose jobs combine information, communication and repetition. Customer support agents are one example. But the same logic can apply to sales teams, marketers, recruiters, analysts, consultants, administrative workers and many managers.
The common pattern is simple: if a job involves turning information into useful communication, AI can often help. It can draft, simplify, compare, summarise and organise. But the human still has to decide whether the output is correct, appropriate and useful. The best workers will not be the ones who let AI think for them. They will be the ones who use AI to remove friction and protect their attention for better judgment.
The Risk: Faster Work Can Become More Work
There is also a darker side to the productivity story. If AI helps workers complete tasks faster, companies may simply raise expectations. A 15% productivity gain can become a 15% higher workload if organisations do not think carefully about how the saved time should be used.
This is already one of the biggest debates around AI at work. Does it give workers more time for higher-value activities, or does it make every day more intense? The answer depends on management choices. AI can reduce stress if it removes repetitive work. But it can increase stress if it only makes the pace faster without changing priorities.
What Companies Should Actually Do Now
The practical lesson is not “give everyone AI and hope productivity rises.” Companies need to identify the tasks where AI can create measurable value. They should start with repetitive information work, customer communication, internal documentation, first drafts, research summaries and operational support. Then they should measure whether the tool actually improves output, quality and time saved.
The second lesson is training. Workers need to know how to use AI well. Bad prompts, weak context and blind trust produce poor results. Good use of AI requires clear instructions, source checking, review habits and an understanding of when human judgment is still essential. The best companies will treat AI as a workflow change, not just a new software tool.
The Real Takeaway
The 15% productivity figure is important because it proves something many people suspected: AI can already make work faster in real situations. But the deeper lesson is more useful. AI creates the biggest advantage when it helps workers handle information, language and repetitive decisions more effectively.
For workers, this means AI literacy is becoming a practical career skill. For companies, it means productivity gains will come less from hype and more from redesigning real workflows. And for everyone else, the message is clear: AI is not just changing what people can do. It is changing how fast ordinary work can move.
For the original research behind the 15% productivity figure, the study Generative AI at Work is the best external reference. And if you want to understand how employees are already using AI in daily work, our article on what workers are using AI for in 2026 is the most natural next read.
