The Role of AI in Enhancing US Business Productivity

We now have quantifiable evidence on the usefulness of generative AI systems such as ChatGPT for real-world business tasks: three new studies examined quite diverse sorts of users in different fields and came to the same conclusion. Productivity rose dramatically, with the least skilled users benefiting the most. Some studies also discovered increases in the quality of work items.

Here, I discuss the conclusions of three research, which are detailed in separate articles

client support representatives handle client inquiries at an enterprise software company. Tutorial 3: Programmers creating a modest software project that took roughly three hours to complete without AI aid. In all three cases, users were measured as they completed the tasks: always for task time, sometimes for quality. Approximately half of the users completed the tasks manually, without the assistance of AI, while the other half used an AI tool. This article presents productivity findings. Is the increase in productivity brought about by AI significant? Productivity for UX Professionals Improving Quality using AI Human-Computer Symbiosis Narrowing Skill Gaps Narrowing Skills Productivity gains were most significant in cognitively demanding tasks, with faster learning and research weaknesses identified. Three Research Studies Productivity Findings The study's most startling finding is that AI is useful in real-world business applications. Users were significantly more efficient at doing their tasks with AI aid than without it. Productivity counts how many tasks a user can do in a certain time frame, such as a day or a week. If an employee can complete twice as much work, their productivity will increase by 100%. These are the results: found that using AI enabled support agents to manage 13.8% more client inquiries per hour. found that using AI enabled business people to write 59% more business documents per hour. AI enabled programmers to develop 126% more projects per week.

The chart below summarizes the conclusions from the three research studies

The measured increase in users' task performance while utilizing generative AI tools (in comparison to control groups of users who did not utilize AI), according to the three research studies covered below In all three cases, the difference with the control group (who conducted the work traditionally, without AI technologies) was statistically significant at p < 0.01 or better. The graphic clearly shows that the decrease in work productivity varies significantly across the three domains tested. It appears that more cognitively demanding tasks (such as developing code versus addressing a customer query) benefited the most from AI aid. Is the increase in productivity brought about by AI significant? In the three trials, generative AI technologies enhanced business users' throughput by 66% when doing realistic jobs. How do we evaluate this number? A number by itself is meaningless. We can only draw conclusions by comparing it to other numbers. According to the Bureau of worker Statistics, average worker productivity growth in the United States was 1.4% per year between 2007 and 2019, prior to the COVID-19 pandemic. According to Eurostat, average worker productivity growth in the European Union was 0.8% per year for the same time. Both figures represent the average value produced by a worker per hour worked. If employees work longer hours or more warm bodies join the workforce, the overall amount of economic output increases, but this does not imply that workers have grown more productive in the sense stated above. This article explores how much value employees generate every work hour. If this value is increased, living standards will rise. We now have something to compare against! The 66% productivity gains from AI are equivalent to 47 years of natural productivity improvements in the United States. And AI translates to 88 years of growth in the European Union, which is one-third greater than the 66 years since the founding of the European Community (the EU's predecessor) in 1957. AI is a huge thing, yes! Caveats

There are three drawbacks to these results.

First, the 66% productivity increases are attributed to previous versions of generative AI (used when the data was obtained), specifically ChatGPT 3.5. We're already working on the next release, which is far better. I expect future AI systems to advance considerably more, especially if they are built based on user-experience inputs rather than being driven solely by engineers. (Current AI tools have significant usability issues.) Second, just one study (customer support) tracked personnel over several months. Studies 2 and 3 (creating business papers and programming) assessed participants' performance during a single usage of the AI tool, which was frequently their first time using AI. There is always a learning curve in utilizing any design, and consumers improve with repeated exposure to the user interface. Thus, I anticipate that the (already extremely high) improvements shown in trials 2 and 3 will be much bigger in real-world contexts where people continue to use tools that make them significantly more effective at their jobs. Third, productivity benefits are only seen when workers execute jobs that benefit from AI support. Many tasks in specific professions, such as UX design, may be unsuitable for AI support, resulting in very minor advantages when examined over the course of their full workday. These factors point in opposing directions. Which will emerge stronger remains to be seen. For the time being, because I do not have any data, I will presume that they are roughly similar in strength. Thus, my preliminary estimate is that adopting generative AI across all business users might boost productivity by roughly 66%.