Is AI a Job Killer? AI-Driven Development Tools Offer Clues
The hand-wringing around artificial intelligence killing off jobs is likely overhyped.
But before we debate the impacts of AI, we have to first agree on what AI even is.
The term “AI,” as it is most often used today, is really just a software machine that analyzes data and predicts what’s likely to happen based on that data. Generative AI, popularized by ChatGPT, identifies patterns in data to create new content, such as software code, essays, images and art.
While it is remarkable to see a text prompt generate content seemingly out of nowhere, it’s just an acceleration of the software automation that’s been underway for more than 30 years. The machine is not creating something wholly new. It responds to your request by interpolating content it has seen in training data and delivering it in a new form.
This might feel like an act of creation, but only in the basest sense. Like an apprentice repeating the efforts of a craftsman, the machine offers back a product that reflects only what we have told it and in a way that we directed it to deliver — potentially hallucinating to fill in gaps for which it does not have sufficient information. Automation has moved well beyond a machine stamping out duplicates, but this is still more diode than Da Vinci.
Why does this nuance matter? Because it gets to the heart of the future of AI and work.
Software Development: The Frontline for AI-Enabled Work
The large language models that power tools like ChatGPT have historically relied on an understanding of the structure and rules of the languages they attempt to emulate. And the languages with the clearest structure and rules? Programming languages.
In addition, it is much harder to review the correctness and accuracy of natural language, whereas code is more easily validated; you can confirm it does what you want as an executable and you can validate it through syntax checks. Programming languages also don’t come with the risks of bias, toxicity and the subtlety of meaning.
Given all of this, it is no surprise that the first knowledge worker role to see generative AI-powered assistants emerge was software development.
My company, Tabnine, created its first AI-based software coding assistant in 2018. This new breed of tools started with the simple ability to predict the code a developer was typing and offer suggestions, but it quickly evolved to generate whole functions, highlight bugs and defects, explain someone else’s code and even generate documentation and tests.
AI coding assistants are now estimated to be in use by over 2.5 million developers. They typically automate 30% to 50% of code generation and cut task completion time for developers in half, as noted by a McKinsey study. With more than five years of history and 5% to 10% of developers using this class of tools, we are starting to see how generative AI can transform a role that was previously perceived as immune to automation.
So is AI killing software development jobs? Not even close.
Boosting Productivity and Satisfaction
Automation from AI-enabled software development tools has thus far reduced low-level tasks and automated mundane and repetitive tasks, enabling developers to do more. There is usually initial doubt among developers in adopting AI assistance, but that doubt typically transforms into an appreciation of the acceleration that AI brings to the role.
Studies have shown that developers who use AI coding assistants report significantly higher job satisfaction and an ability to stay “in flow” longer, and they have greater mental energy to apply toward more strenuous tasks. They also report being able to complete tasks that they would have failed at before – as much as 10% more.
This research potentially points to a pattern that looks to be emerging around AI tools augmenting knowledge workers in all kinds of jobs and industries. AI automation’s ability to help workers shift time from low-level grunt work to higher-order thinking and creativity can create greater value for employers and more meaningful work for workers.
As Turing Award-winning computer scientist Donald Knuth stated, “Some tasks are best done by machine, while others are best done by human insight; and a properly designed system will find the right balance.”
This new level of automation could not come at a better time. Job growth for software developers, quality assurance analysts and testers is projected to grow 25% from 2022 to 2032, the U.S. government says. No way we will produce 25% more skilled workers to fill those jobs in the same timeframe — or even get them from other countries.
A survey by GBK Collective found that 78% of companies expect to use AI for software development within three to five years. Freshworks estimates that U.S. companies could save over $15,000 per IT employee each year with AI automating repetitive tasks. An IDC study found that nearly 40% of IT executives said generative AI “will allow us to create much more innovative software.” Contrary to replacing jobs, large enterprises reported to IDC that they expect AI to help them overcome the shortage of skilled developers and operators.
Bottom line: AI as machine-driven automation increases productivity, which drives more innovation and, ideally, stronger economies. The kid who worked the neighborhood with a push lawnmower can now work two neighborhoods with an electric one. More earnings. More spending.
This productivity boost is desperately needed, as the shortfall of workers in tech has been mirrored in countless other roles. We need automation to continue enjoying the comforts of modern society and to grow GDP — it’s the only way to meet the labor force needs as boomers retire and birth rates decline.
AI and the Act of Creation
The current advancements in AI differ from automation in the past, in which machines replaced human work that was often physical. This time, AI (particularly Generative AI) will replace more white-collar work. Some jobs will go but others will be created in their place.
As the McKinsey Global Institute states, “We see generative AI enhancing the way STEM, creative, and business and legal professionals work rather than eliminating a significant number of jobs outright. Automation’s biggest effects are likely to hit other job categories. Office support, customer service and food service employment could continue to decline.”
Of course, it hurts individuals when jobs go away. Certain types of workers will suffer more. Then again, surprises occur when technology changes the world. When ATMs arrived, everyone predicted the death of the bank teller. Not so. As Reddit explains, the rise of ATMs meant fewer tellers per branch. But because it became cheaper to open branches, more branches arose and so did teller jobs.
The roles that will continue to be in high demand are those rooted in the act of creation. And not just artists and writers, but anyone who knows how to turn an idea into something new. As we learned in software development, the machine can show you options, fill in the gaps and point you toward the right answers, but it can not independently create a new application or business. AI lacks both ideas and innovation. Use GenAI well, and it’s like putting on a suit that amplifies your speed and agility. Use AI to create something you do not have a vision for, nor truly understand, and it will create gibberish.
Humans are Different
Humans and machines differ, and they always will.
“The question of whether AI will replace human workers assumes that AI and humans have the same qualities and abilities — but, in reality, they don’t. AI-based machines are fast, more accurate and consistently rational, but they aren’t intuitive, emotional or culturally sensitive. And, it’s exactly these abilities that humans possess and which make us effective,” writes AI expert, author and professor David De Cremer and Garry Kasparov, chairman of the Human Rights Foundation, in the Harvard Business Review.
The authors conclude that AI will augment human intelligence, not replace it. I couldn’t agree more, and that’s why I think hand-wringing should really be replaced by high-fives.