In March, Ethan Mollick, an associate professor at the Wharton School of the University of Pennsylvania, conducted an experiment. Mollick, who studies innovation and entrepreneurship, wondered how much he could accomplish in a business project in half an hour. The goal: to use artificial intelligence tools to create promotional materials for the launch of a new educational game. The AI would generate all the resources, and Mollick would be the guide.
Using AI to Boost Productivity
Mollick used Microsoft’s Bing Chat, powered by the GPT-4 model, to gather information about his product and its market landscape. He also created an email campaign and social media posts for the launch and devised a plan for a webpage to promote the game. Then, Mollick opened a separate tab, and GPT-4 helped him create HTML formatting to build the actual page. MidJourney, an AI program that generates images from written prompts, provided Mollick with a large and attractive image to welcome visitors to the page. Finally, he produced a promotional video. Again, Bing generated the script. Then Eleven Labs, a program that helps develop natural-sounding speech, took all the components and crafted them into a video.
Time was up. Mollick had enough content to go live.
The Potential Impact of AI on Productivity
At its core, productivity is a dull mathematical equation that compares worker output over a specific unit of time. The more tasks a worker can complete in one hour, the more productive they are considered.
By this definition, AI undoubtedly made Mollick more productive. After all, how many companies can build a promotional campaign in thirty minutes using just human power and intellect? How many hours were saved thanks to these smart technologies? How many knowledge worker tasks were completed? These questions regarding the productivity of knowledge workers are now at the forefront of interest for economists, companies, researchers, and managers pondering the impact of AI.
If productivity is equivalent to economic growth, it seems that certain AI tools will enhance productivity. A recent report by McKinsey Global Consulting projected that AI could inject $4.4 trillion into the global economy each year. Banks alone could generate between $200 to $340 billion more in customer service, decision-making, and fraud tracking using AI. The report estimated that healthcare and pharmaceutical companies could see a 25% increase in profits if AI assists in developing new drugs and medical materials.
Studies conducted by Capgemini Research Institute, a research center focusing on the impact of digital technology, indicate that productive AI will also boost productivity in sectors such as IT, sales, and marketing.
The Impact of AI on Knowledge Workers
While the accessibility and spread of AI tools will generally increase productivity for companies, research has also found that AI tools and automation have an effect on how knowledge workers focus and get work done. In a study conducted by The Economist under the auspices of Dropbox, the Economist Impact found that 79% of those using automation tools, including AI, in their work are more productive, while about 70% say they are more organized. Another report by McKinsey estimated that AI could save 30% of work hours in the U.S. for knowledge workers by assisting them with programming, responding to emails, and managing other routine tasks.
However, Bhaskar Chakravorti, an economist and consultant at the Fletcher School at Tufts University, who has been using AI to help him complete emails automatically, questions how much one tool can really accomplish. He said, “When the emails start piling up, I find myself hitting the tab button and completing the sentence suggested to me by Outlook,” adding, “Does this little extra time I saved on the last email really lead to a significant increase in my productivity? I don’t think so.”
Ultimately,
Other, just because we can use artificial intelligence tools to do more in less time does not necessarily mean that the results will be better. Experts argue that the question is not how quickly AI can get us to the finish line, but how it can be used to empower critical thinking, creativity, and focus more for knowledge workers. If we are able to delegate mechanical tasks in our to-do lists to AI, how does that affect how we spend work hours and ultimately measure human productivity?
How AI Can Affect Productivity as We Know It
The question of AI’s impact on knowledge work is not new. Researchers have been studying the effects of this technology, and so far, the results from early studies look promising.
In a study conducted in 2022, Github recruited 95 software developers to find out how long it would take them to write an HTTP service in JavaScript. (All participants were professionals, so this would be a task everyone could complete without assistance from tools.) Half of them were allowed to use GitHub Copilot, an AI-based tool that can autocomplete code. The programmers who used Copilot completed the task 55% faster than those who did not use it.
Recently, researchers at the Massachusetts Institute of Technology measured how ChatGPT could affect productivity in writing tasks. 444 college-educated professionals were asked to complete writing tasks with and without the assistance of ChatGPT, and the final reports were evaluated by professionals. ChatGPT reduced the time participants spent on the task by improving the quality of their work.
These early findings suggest that much of the tedious and time-consuming work on to-do lists can be delegated to AI. So what remains on the rest of the list? The authors of the MIT study suggest that tools like ChatGPT can help writers focus more on idea generation and editing. The Github Copilot study found that with more free time, developers could focus on more fulfilling work and ultimately find greater enjoyment in the programming they do. The underlying idea is that AI can help support the tasks of knowledge workers – like scheduling calendars and organizing emails.
However, linking more free time to increased productivity is difficult. Although these studies reflect the potential for AI to have a positive impact on productivity, critics have pointed out that the classical definition of productivity is not sufficient to evaluate the productivity of knowledge workers. This is because knowledge workers create value for their companies and organizations through what they produce, rather than how quickly they produce it. The outcomes depend on a much larger scope based on the group of workers in the organization rather than the individual. And let’s not forget that companies do not necessarily track working hours for salaried employees, meaning that worker productivity isn’t as simple as solving an equation. Company results may be directly measurable in terms of productivity, according to Chakravorti. “But in some cases, it is not.”
Ultimately, initial research finds it easy to measure the efficiency of AI in synthesizing mechanical tasks, but the larger effects for knowledge workers – more free time to generate ideas and focus on hard work – are difficult to track and define. Searching for the right metrics
The Right Metrics for Measuring Productivity
For some people, when it comes to expanding the future of knowledge work, productivity as we know it is not the right metric at all. Dan Schreiber, CEO of Every, a newsletter covering technology, AI, and productivity, says, “How many page views were generated? How many emails were increased? What are our revenues? We do not look at hours worked or number of words written,” and adds, “These metrics do not necessarily align with just productivity and are not problems that can be easily delegated to AI. What it requires is a mix of something else entirely – complex thinking, critical problem-solving, and the time and mental space required for creative output.”
He says:
“For knowledge workers who naturally work in a creative industry, productivity and creativity are in conflict with each other.” “Creativity is an unpredictable magical process where you literally try to extract something from your soul,” he adds. “Productivity requires being on a schedule and reliable. This is really tough.”
If the primary tools of artificial intelligence enable us to do more than humans alone can, thinking critically about productivity when it comes to AI technology does not quite capture the whole picture. Chakravorti says, “These studies are not only limited in the points being measured, but are often limited by industry as well.” More research needs to be done on knowledge workers in various fields before we have enough data to say exactly how AI could impact work in different areas.
Changing the Face of Productivity
Currently, no one really knows how to measure the real impact that AI will have on knowledge workers, whether through productivity or something entirely different.
With tools like GPT, the impact on traditional productivity – simply getting tasks done – is quite evident. Chakravorti says, “It’s clear that people are writing white papers, articles, marketing, graphics, and infographics faster than they were in the past.” However, the outputs provided by generative AI seem relatively limited when it comes to the quality of creative output. (As a journalist, I’ve sometimes used GPT-4 for research, and while it reduces my time spent sifting through Google search results, I typically end up spending more time critically thinking about the suggestions and sources the program proposes.)
When considering how to free up more time for knowledge workers to engage in more creative and complex tasks, our way of thinking about human productivity – and work in general – can take a different direction. Shepper says, “A lot of knowledge work might shift to resemble much more what managers do instead of what entry-level employees do. Everyone is moving up to some extent.”
While traditional definitions of productivity still hold, companies will need to think of alternative ways to measure their workers’ performance, or focus on larger metrics as Shepper does at Every. It’s easy to imagine a future where a team of analysts at a consulting company relies on using GPT to prepare a report. The work that previously required a manager and two analysts might now need just one automated tool. Then, individuals freed from this initial report can go on and do something completely different and have more time and energy for higher-level tasks.
He says, “When you can accomplish everything [Molek] can do in thirty minutes, those skills will be very important: vision, taste, and the ability to prioritize.”
Leave a Reply