Summary: Two research studies show that software developers who use AI-powered tools when programming experience greater job satisfaction and creative fulfillment. They also felt that the AI support helped them stay in flow and preserve mental effort. At least in this domain, the emotional impact of AI is highly positive.
We know from previous research that AI can boost developer productivity tremendously. In a study I discussed in my article about AI-driven productivity, programmers produced 126% more per unit of time when assisted by the AI-driven GitHub Copilot tool. Developers experienced the largest productivity gains of any occupation in this research. My theory (which awaits further research for validation) is that the more cognitively demanding the job, the bigger the gains reaped from cognitive tools: AI is a forklift for the mind allowing users to offload mental burdens and focus on higher-order thinking.
Increasing productivity by 126% means that an employee can complete more than double the amount of work. This is fantastic for business! But it’s fair to ask whether producing more output is necessarily good for the employee. Maybe they’ll score a bigger bonus, but will they feel overworked or burnt out?
Fortunately, two research studies address this latter question. And the findings are promising.
In the first study, Eirini Kalliamvakou from GitHub reported on a study of over 2,000 developers using their AI tool, Copilot. The fact that the company itself did the study makes me wish for more independent research. Still, since it’s those 2,000 outside developers (not working at the company) who are speaking to us through their responses, I find the findings credible. (I refer to this study as the “GitHub study” below.)
In the second study, Begum Karaci Deniz and colleagues from McKinsey reported on a study of over 40 McKinsey developers performing various programming tasks with or without AI assistance. (I refer to this study as the “McKinsey study” below.)
Results from the GitHub Study: Developers Feel More Productive and Happier with AI
The first finding is not surprising, given what we already know. 88% of the study’s developers said they were more productive using the AI tool. We also know from the research I reported previously that programmers tended to underestimate their own productivity gains from AI, perhaps because the increases seem so implausibly large that even people’s own experience doesn’t allow them to grasp the gains fully. Thus, even though 88% is already a towering figure, I suspect the true number is even loftier. (As always, it’s a problem with self-reported data that it’s not always entirely accurate relative to objective measures.)
The other measures related to each developer’s personal feelings, and here I will accept the veracity of self-reported data. If people profess happiness, I agree that they are happy.
The satisfaction scores with the AI tool were very high. 60% of developers reported feeling more fulfilled with their job. This alleviates my concerns that employees might resent producing more work.
Even more striking, 74% said that the AI tool allowed them to focus on more satisfying work, likely because the AI handles the grunt work of coding. Finally, 59% said that they felt less frustrated while coding.
“Happy programmer” generated by Leonardo.AI using Chinese ink painting style.
These are excellent scores for improved worker satisfaction with new technology. It is more common to witness disgruntlement when a new digital tool is introduced in the workplace. (If you ever ask a physician about his or her experience with a new health record system, brace yourself for an earful of grievances. I have yet to meet a doctor who doesn’t harbor disdain for these modern clinical computer systems.)
One user, quoted in Eirini Kalliamvakou’s report, said:
“[With Copilot] I have to think less, and when I have to think it’s the fun stuff. It sets off a little spark that makes coding more fun and more efficient.”
This anecdote finds validation in the full study group of 2,000 developers. A staggering 96% said that Copilot made repetitive tasks faster. No intelligent person relishes repetitive tasks, and having AI shoulder them is likely a big reason that the programmers feel more fulfilled now that they can focus on more interesting concerns. (It’s almost unheard of to get 96% of users to agree with anything in user research.)
87% reported expending less mental effort on those repetitive tasks. Since I’m discussing the emotional impact of AI rather than the business impact, this number may be even more critical than the speed number. The sensation of a liberated mind is exhilarating.
Results from the McKinsey Study: AI Use Makes for Happy and Satisfied Developers
Similar to the GitHub study, the McKinsey developers were happy about using AI tools. 88% agreed to feeling happy when programming with AI assistance, compared to 45% who felt happy programming without these tools.
As another replicated finding, the McKinsey developers also felt that they could focus on more satisfying and meaningful work when assisted by the AI tools. 87% agreed to this statement when using the AI tools, compared to 50% who agreed when programming without AI assistance.
Two Studies, Same Findings = More Credibility
The two studies discussed in this article have roughly the same findings. Yes, the exact percentages differ, as would be expected: the two studies included differently selected groups of participants who performed different tasks while using different AI tools. (McKinsey doesn’t name the tools used in their study.) And the two studies were obviously conducted by two different teams of researchers, working in different companies.
The difference between the studies is exactly what makes it so exciting that the conclusions are the same: more happiness, more meaningful work, more flow. Research credibility shoots through the roof when the same conclusions are found under vastly variant conditions.
(It’s too common for a single study to turn out to be wrong, because of some obscure flaw in the experiment. Of course, two independent studies can in theory share the same flaw, but this is unlikely.)
More Flow = More Bliss
The most important finding about AI’s impact on the users’ emotional well-being is that 73% of developers reported being more in the state of flow when using AI in the GitHub study.
In the McKinsey study, an even more impressive 94% of developers agreed to feeling in a flow state when programming with the AI tools. This should be compared to 55% of developers saying the same about programming without AI assistance.
Flow, a concept developed by psychologist Mihaly Csikszentmihalyi, is a mental state of complete immersion and focus in an activity. Being in flow results in a loss of self-consciousness and a sense that time passes quickly. Flow is achieved when a person engages in an activity that challenges their skills, requires focus and concentration, has clear goals, and provides immediate feedback. In this state, people become deeply involved in the activity for its own sake, rather than for external rewards. Flow is often associated with heightened enjoyment and satisfaction in work or activities.
Thus, 73% of users feeling more in the flow (GitHub) or 94% of users feeling in flow (McKinsey) when using AI is a solid testament to why this tool improves mental wellbeing. This research suggests that for software engineers, AI brings not only greater productivity but also greater joy. Let's hope the same applies to AI use in other professions. I eagerly await those research results, but based on the current data, the future of human-AI symbiosis seems bright. Properly designed, AI will not dehumanize and diminish workers but rather elevate and actualize them.
Geek Triumphant, Thanks to AI. (Generated by Midjourney from the prompt “Happy Geek”, but I decided to retitle the image for use in the conclusion.) Sorry for the stereotype, my fellow geeks, but this is what the AI gave me. I do wear glasses myself, so there's that. Nothing said about my biceps.)