Feature
A recent study looked at how much time AI saved employees overall. The result? Less than 15 minutes a day.
Ever since artificial intelligence hit the mainstream in 2022, vendors and enterprises alike have justified the enormous investments required by talking about how much time employees would save.
But a survey from the National Bureau of Economic Research trying to quantify these gains across companies found just 2.8% improvement — 13.44 minutes across an eight-hour day.
What happened?
Table of Contents
- Experts Split: Is This Productivity Dip a Warning or a Pattern?
- The Truth Behind 'AI-Driven' Job Cuts
- The Paradox That Predicted Today’s AI Plateau
- The J-Curve Theory: AI Gains Start Slow and Jump Later
- What Makes These Findings Unique?
- 4 Ways to Actually Move the AI Productivity Needle
- What the Next Wave of Research Could Reveal
Experts Split: Is This Productivity Dip a Warning or a Pattern?
It's surprising that boosts in improved earnings or changed hours have not popped up yet, given how fast AI adoption has been, noted Anders Humlum, assistant professor of economics at the University of Chicago’s Booth School of Business and conductor of the study.“My expectations before I looked at the data was that tools would already have made an impact.”
So what reaction has the study gotten? According to Humlum, there are two camps of reaction:
- Surprised at the lack of quantifiable gains; and
- Not surprised due to a historical perspective.
"Even the most transformative technologies took decades to really make a difference for the economy," Humlum noted about the latter group. "If you take that lens, it’s less shocking that these tools have not upended the economy.”
Related Article: The AI Productivity Paradox: Why I'm Working More and Loving It
The Truth Behind 'AI-Driven' Job Cuts
“One should be a little careful connecting everything to AI. Just because some firms are shrinking doesn’t mean AI has driven all of this."
- Anders Humlum
Assistant Professor of Economics, University of Chicago
But AI has upended the economy, hasn't it? Companies ranging from Amazon to Microsoft have laid off thousands of people, saying their jobs could now be performed by AI.
Still, said Humlum, “One should be a little careful connecting everything to AI. Just because some firms are shrinking doesn’t mean AI has driven all of this. Amazon, in particular, had a huge expansion during the pandemic, and now that we’re out of the pandemic, it makes sense that these firms would go back to trend a little bit.”
There's also a pattern to these layoffs, noted Enrique Dans, professor of innovation at IE University in Spain.
"Many cuts are episodic, strategy-driven and often paired with aggressive AI capex, rather than the direct, one-for-one substitution you’d expect if GenAI were immediately removing vast quantities of labor," he explained, adding that the system can absorb a lot of task-level automation through reassignment and hiring slowdowns while still exhibiting high-profile trims in specific companies and functions.
"Both the ‘quiet churn’ and the ‘loud cuts’ are part of the same transition.”
The Paradox That Predicted Today’s AI Plateau
Experts say these results are actually normal for technology adoption. As American economist Robert Solow put it: "You can see the computer age everywhere but in the productivity statistics."
Dans noted, “Economists then spent a decade showing why: diffusion lags, mismeasurement (especially of quality and free digital services) and the need for complementary intangibles (process redesign, training, data, integration) that don’t show up immediately in GDP.”
The J-Curve Theory: AI Gains Start Slow and Jump Later
In the 2010s, this concept was refined into the Productivity J-curve, Dans said. “Early years of a general-purpose technology (like AI) understate productivity because firms are investing in intangibles and reorganizing; the measurable payoff comes later, sometimes suddenly. That conceptual toolkit maps cleanly to today’s generative-AI moment.”
So what’s next?
“If you take the J-curve seriously, the window for visible aggregate gains is mid- to late-decade as diffusion deepens and firms industrialize use cases beyond pilots, but it will be uneven, with customer-support, back office and coding workflows leading and physical-world sectors lagging,” Dans said.
“Expect the Danish pattern — reorganization first, macro stats later — to be a decent preview for advanced economies with high adoption and strong internal change management.”
Related Article: How I Use AI to Boost My Personal Productivity
What Makes These Findings Unique?
To a certain extent, this survey could be seen as an outlier. Other studies have shown as much as a 15% improvement in productivity due to AI tools.
Denmark’s Data Advantage
One difference is the methodology, which was set in Denmark. That allowed researchers to link survey responses to administrative data to confirm figures such as earnings, employment, wages and hours of work.
“The key advantage is that allows us to get a more objective measure about whether these tools are making an impact,” said Humlum. “If I go to a manager and ask, he may say yes, but who knows. We wanted something more objective than people’s perceived impact.”
A Labor Market That Blunts the Impact of AI
The survey's results could also have been affected by the Danish work model, Dans said, noting that Danish employment policies such as easy hiring and firing for firms, generous unemployment insurance, high union density and coordinated wage-setting tend to absorb shocks or improved efficiency as role reassignments and internal task churn, rather than across-the-board layoffs or wage spikes.
“The NBER result does not say ‘AI doesn’t work,’” Dans said. “It says that in a high-adoption, high-coordination labor market like Denmark’s, early gains are absorbed as task reallocation and job redesign rather than as measurable wage or hour changes."
Webinar
Dec
2
From Manual to Magical: How AI Transforms CX Teams
Learn how to replace manual support processes with automation that actually delivers.
Register
Webinar
Dec
9
Rebrand. Migrate. Optimize. How to Do It All (Without Slowing Down)
Cresta leveled up site speed, design flexibility and marketer sanity (in record time). Find out how.
Register
Webinar
On demand
How to Build a Solid Knowledge Foundation for AI Success
See how leading brands keep their AI honest, compliant and actually helpful.
Watch Now
Webinar
On demand
Fix the Content Bottleneck: Build a Better WebOps Strategy
Content stalled? Dev overloaded? You’re not the only one. Learn how streamlined WebOps bridges the publishing gap.
Watch Now
Webinar
On demand
Beyond Storage: Smarter Content, Bigger Impact with DAM + AI
Discover how the DAM + AI duo makes content smarter, stronger and more accessible.
Watch Now
Webinar
On demand
Agentic AI Playbook: Real-World Customer Service Use Cases You Can Deploy Now
Boost self-service by 30% and slash call volume by 63% with agentic AI.
Watch Now
Transpose that to the US, Dans added, and you’ll see more visible churn, especially where management treats AI as a cost-takeout lever.
Related Article: Do AI Coding Tools Really Increase Developer Productivity? Studies Say No
4 Ways to Actually Move the AI Productivity Needle
All this stipulated, the NBER study found basic steps improved employee productivity using AI.
1. Encourage AI use
One step to improve employee productivity is to encourage the use of AI tools, said Humlum.
“There’s a huge variation within workplaces whether employees are using it. In some workplaces, it’s explicitly expected that workers are using it. If you’re not using the tools, you’re not going to benefit from them.”
2. Implement AI Training
AI training and upskilling programs also matter, Humlum said, including:
- How to get started: AI tool use cases, tools to check out, etc.
- Dos and don'ts: Fact-checking outputs, protecting confidential data, etc.
3. Personalize AI Systems
Enterprises are now personalizing their AI systems rather than using generic solutions, and protecting their proprietary data, Humlum said.
For instance, some employers use a custom version of ChatGPT tailored to their specific workplace. It can provide access to in-house documents but doesn't link confidential data to the provider.
4. Include AI in Workflows
Incorporating AI into workflows will also make a difference, Dans said. “The aggregate productivity needle may lag until firms finish the unglamorous work of rewiring processes."
What the Next Wave of Research Could Reveal
Humlum said he plans to keep studying AI productivity using the same methodology to see how things change as AI becomes more widely adopted. He’s preparing to run a new survey later this year to update the results.
“This is still early days,” he said. “Who knows what is going to happen in ten years’ time?”
