A company selling its wares to a wide variety of customers decides to introduce its customer-support staff to a new support tool based on generative artificial intelligence (GenAI). The employees manning the phone lines or responding to customer complaints over computers now have a new type of assistant to help them in their jobs.
The productivity of these employees is measured by how many complaints get resolved in an hour. The GenAI tool helped increase employee productivity through three effects: a reduction in the time an employee now spends on each customer query, an increase in the number of chats an employee handles every hour, and a higher success rate of cases that are resolved to the satisfaction of callers.
Also Read: Arun Maira: The impact of AI on Indian jobs is a distraction we can do without
Customers also tend to be more satisfied by their experience during the call or chat, which is especially important, since the ones reaching out are usually doing so because they are dissatisfied with the product they have paid money to buy. Less stress from irate customers also reduces attrition in the company’s workforce.
However, the most interesting result of this study by Erik Brynjolfsson, Danielle Li and Lindsey Raymond, based on their observations of 5,179 customer-support agents at work, has to do with the skills gap in the unit they put under their research microscope. The three economists found that while the overall productivity of the team increased, most of that improvement was among team members who were either less skilled at their jobs or less experienced.
In other words, access to the GenAI tool narrowed the gap between the best and the rest. And there was durable learning by the latter. Their performance continued to be better than before even when the tool was not available to them later—say, when there is a software outage that prevents them from getting assistance from the GenAI tool.
Also Read: DeepSeek’s breakthrough is a pivotal moment for the democratization of AI
The rise of artificial intelligence has raised all sorts of questions about its impact on economies, societies and the future of work. For example, will it throw millions out of jobs? Or will it lead to a new golden age of economic growth by driving productivity to new heights?
After the launch of the Chinese AI model DeepSeek-R1 in January, Olivier Blanchard, a professor of economics at the Massachusetts Institute of Technology, described it as “probably the largest possible TFP shock in the history of the world.” TFP, or total factor productivity, is a measure of how efficiently a society uses its labour power and capital assets.
The importance of these macro issues—both positive hopes as well as negative concerns—should not be underplayed. However, some studies, such as the one mentioned earlier in this column, also point to the possibility of how AI can help narrow skills gaps within an organization or among a wider population doing some specific task.
There could be other effects as well. Think of a colleague who writes very well but cannot code or another who is a coding whiz but struggles to write a note to the senior management. Will both be on equal ground in the near future?
Also Read: Rahul Jacob: DeepSeek’s big AI shake-up holds policy lessons for India
As with the case of the customer-support team, another piece of research in Japan offers a similar story of how access to AI narrows the productivity gap among taxi drivers in a Japanese city.
This study by Kyoto Kanazawa, Daiji Kataguchi, Hitoshi Shigeoka and Yasutora Watanabe shows that taxi drivers in Yokohama shortened their cruising time when they got access to an AI support system to find customers, and that it was low-skilled drivers who captured most of the benefits from better demand forecasting. (Yokohama did not permit services such as Uber and Grab when this research was conducted by the team of economists.)
However, one of the most encouraging results in the context of countries such as India was recently reported by a World Bank team in Nigeria. For six weeks in June and July 2024, secondary school students, 800 in all, were given after-school English classes in a computer lab twice a week. They were guided by teachers on how to use a GenAI tool to improve their English. Students who were randomly chosen to participate in this programme later outperformed their peers who did not attend these after-school sessions.
While the initial results reported in a recent blog by the World Bank team show that “generative AI, when implemented thoughtfully with teacher support, can function effectively as a virtual tutor,” the distribution of test scores among the “treatment group”—or the students chosen for this study—also shows a more even spread than in the peer group.
Also Read: The silent crisis: Pandemic learning losses that could haunt a generation
India has long been struggling with the quality of education provided in its school system. There are many underlying causes, but the results of the Nigeria study highlight the possibility of some form of AI helping students who are lagging either because of poor teacher quality or social disadvantages that make it harder to overcome hurdles in their way.
Can GenAI provide the sort of high-quality tutoring after school that children from elite families get?
The case for AI as a way to close an organizational or societal skills gap leads to a question with profound implications. Will a society where everyone is an expert also be a society where nobody is an expert?
The author is executive director at Artha India Research Advisors.
