We’re still a ways off from the artificial intelligence (AI) seen in movies like Bicentennial Man and I, Robot since ChatGPT launched in December 2022, it feels like we’ve been experiencing continuous breakthroughs in its capabilities. When it comes to software development, there are countless articles about how GenAI is transforming the field, but much of it is anecdotal or theoretical.
We wanted to find out the reality of GenAI in software development today: its prevalence, applications, effectiveness, and more.
At BairesDev, one of our greatest resources is our thousands-strong engineering team. They’re on the front lines of software development for hundreds of clients, ranging from Fortune 500 companies to nascent startups. What better way to get a handle on the use of GenAI in software development than to simply ask them about their use of it?
We received survey responses from 500+ software engineers of varying experience levels and found:
GenAI Is a Key Component in the Software Development Process
- 72% of engineers are leveraging GenAI in their software development process.
- Almost half (48%) of them are using GenAI every single day.
- 81% are using GenAI to write code they used to write manually.
- Still, 40% of engineers do not believe GenAI has freed up time for them to accomplish other tasks.
A recent Thomson Reuters survey found that just 12% of white-collar workers are currently using GenAI, and 11% have active plans to use it. The remainder are still considering it or have no intention of using it. This survey spanned respondents in professional services industries, including legal, tax and accounting, risk and fraud, and government professions. When it comes to software engineers, they are taking full advantage.
It’s impressive how far we’ve come in such a short amount of time. Just two years ago, most developers were not leveraging AI at all, and now the vast majority are. Not only that but of the engineers who are using it, 87% are either using it every single day or at least several times per week.
Looking at the chart below, we have to talk about code generation. When engineers are using GenAI to execute on this, it is for net-new code snippets. AI is not great at creating full systems or piecing together existing code. It is great at “scaffolding” and creating starting pieces of code snippets to be edited and “glued together” by the engineer. Think of it like building a car. AI might be able to generate a transmission, but it wouldn’t know how to integrate the transmission into the engine.
Source: BairesDev
GenAI Is Giving Engineers a Productivity Boost
It appears the pace at which products can be built has clearly increased because since adopting GenAI, the majority of engineers have reported significant increases in productivity. 23% of the survey population’s GenAI users reported a productivity increase of a whopping 50% or more. 71% believe their productivity has increased between 10%-25%. Only 6% of engineers reported having “no change” in their productivity since they started using GenAI.
The roles experiencing the highest productivity boosts from GenAI are Site Reliability Engineers, DevOps, GIS Developers, and Project Managers/Scrum Masters (all 40%-50+%). On average, Data Scientists report a 32% increase while Full-Stack Developers believe GenAI has enabled their productivity to increase by 27%.
When it comes to quality, 74% of engineers say GenAI has increased their quality of work to some extent. 24% believe there has been no change and only 2% believe the quality of their work decreases when leveraging GenAI. More than half of engineers (53%) say the quality of their work has improved between 10%-25%.
Similarly to how Grammarly leverages AI to provide suggestions and best practices, engineers are able to use similar tools to improve code quality. The suggestions do not always work, but often enough, they help improve code quality and efficiency.
Software engineers are developing a symbiotic relationship with AI as its capabilities improve. This will lead to efficient workflows and higher-quality software. GenAI will be similar to search engines like Google: We depend on them in our daily activities, but it’s a tool to improve efficiency that still requires human supervision.
Software Engineers Are Turning into Editors
Just like a human, AI-generated code will have errors. In reviewing their AI-generated code, 47% of engineers report errors every time, although typically minor. Another 16% say there are errors every time, but they are typically significant. Either way, 63% find errors every single time the AI generates code. We are far off from perfection, but workers are now able to take an editor’s approach to writing code instead of being confined to repetitive coding tasks.
Source: BairesDev
For engineers with 8+ years of experience, 49% have seen minor errors every time they use AI-generated code, while only 39% of engineers with less experience have seen these errors. Just like with writing code manually, error identification in code can be related to experience and seniority.
Source: BairesDev
What Is AI Not Good at Doing for Software Development?
Get ready for a curveball. The answers varied greatly among our survey respondents, but 20% think AI is not good at code generation. Why would so many engineers be using GenAI for something they believe is not good at executing? Speed. Even though it is not good at generating useable code, it’s able to generate it so fast that is makes it worth it. As I stated above, GenAI is elevating engineers from “laborers” to “managers.” For a chef, it’s easier to tweak a recipe to their desired taste than to devise a whole new dish altogether.
What have we learned? It feels like more than any other vocation, software engineers are taking advantage of GenAI for their core functions even though the technology is far from perfect. GenAI has quickly transitioned the physical nature of the work engineers are doing, and with that comes an additional set of skills that the best talent will possess.
As GenAI takes over repetitive tasks, creativity, problem-solving, critical thinking, and communication skills will become sought-after skills in software engineers. Instead of memorizing commands, devs will collaborate with one another to solve increasingly complex problems. The role will evolve from what we know it today to becoming one of a project orchestrator.
