AI coding assistants wave goodbye to junior developers

Despite mixed early returns, the outcome appears evident: Generative AI coding assistants will remake how software development teams are assembled, with QA and junior developer jobs at risk.

As AI assistants become better at writing code, CIOs and dev leaders will reshape their teams, focusing on AI specialists and senior developers to oversee AI-generated code, some IT leaders say.

Application development teams will be leaner, with the remaining senior developers focused on the best ways to translate product needs into software development, says Anna Demeo, founder of Climate Tech Strategic Advisors and former head of the dev team at Fermata Energy, a vehicle-to-grid application provider.

Companies leaning on AI coding assistants will hire fewer junior developers, interns, and in some cases, product managers, as AI takes over those roles, she says. “When you have big teams, you always have A players and B players, and hopefully not C players, but they exist,” she says. “AI, in some ways, makes it harder to be a C or a B player.”

The remaining developers will need to be critical thinkers who understand the business needs and can work in cross-functional teams with product specialists, the marketing department, and other employees.

Developers as editors

Demeo already sees some client companies reorganizing their dev teams around AI, with senior developers or software architects overseeing and tweaking AI-generated code, she says. She compared the change impacting various roles to the process of publishing a novel.

“Coders no longer have to be writers — they’re editors,” she says. “These senior developers have to understand the content and who the reader is, and in that case, who the customer is and what we’re trying to achieve.”

Future dev teams will be made up of a product manager or business analyst, a UX designer, and a software architect who uses AI tools to generate prototypes, then tweaks the code until it’s ready to ship, adds David Brooks, senior vice president for evangelism at Copado, provider of a DevOps platform for Salesforce.

AI will handle the rest of the software development roles, including security and compliance reviews, he predicts.

“At some point, current software development jobs will be eliminated; junior software developers will be the first to go,” he adds. “Software architects will do less coding and more high-level system design along with keeping an eye on the solution generated by the AI.”

There will be some bumps in the road, Brooks says. The biggest challenge will be training the next generation of software architects — with fewer junior dev jobs, there won’t be a natural apprenticeship to more senior roles.

Coding assistants now ubiquitous

It’s unclear how soon the reshaping of dev teams will reach critical mass, but the use of AI coding assistants is already widespread among developers, according to a recent survey by GitHub. More than 97% of developers from four countries said they have used AI coding tools at work, reinforcing industry observations that coding assistants are among the more popular use cases for gen AI today.

GitHub reported 1.3 million users of its Copilot coding assistant at the end of January, up 30% from the previous fiscal quarter. More than 77,000 organizations have adopted Copilot as of late July, according to GitHub owner Microsoft.

Meanwhile, about three-quarters of IT professionals fear that AI will make their skills obsolete, according to a recent survey by Pluralsight, an online education provider.

Some observers see the impact of AI happening over the long term, with many dev teams ramping up to make the best use of AI in the coming months.

In the next one to two years, dev teams may actually get larger, as additional coaches are needed to improve productivity and build AI prompt engineering skills in existing teams, says Ed Watal, founder and principal at Intellibus, an IT consulting and services provider.

But dev teams are likely to get smaller over the long term, as three software engineers will be able to deliver the code that five or six did in the past, he adds.

At the same time, traditional dev teams will be disrupted, with more employees able to write applications using AI and low-code/no-code tools, Watal says.

“They have the power to write code even though they may not deeply understand how the AI-generated code works,” he says.

While many IT leaders predict that AI coding assistants will ultimately lead to fewer developer jobs, others question the wisdom of turning over most programming to AIs. Some dev leaders questioned the wisdom of using AI to both write and debug code.

Are the benefits overhyped?

Some organizations may have overestimated the time saved with AI coding assistants, says Marcus Merrell, principal test strategist at Sauce Labs, a provider of code testing solutions. A potential 30% increase in developer productivity is a good start, he says, but not a fundamental shift.

“What I’m actually seeing is that teams think they will get enormous benefits from these tools, so they overinvest in tooling, over-rotate on structural and process changes, or overdo the staff reductions they already had planned, based on the gains they imagine they’ll get from the vendors of AI tooling,” he adds.

Merrell doesn’t believe generative AI will replace developer jobs; instead, low-code/no-code tools will have a bigger impact. AI coding experiments will continue with modest success, but eventually, the big AI companies will need to get returns on their massive investments.

“We’re going to spend two to three more years trying to squeeze productivity and magic out of this technology, and then be very slow to admit that it was all a shell game,” Merrell says. “What worries me is that we’ll get hooked on these tools, and then those companies will start to charge the real price that it costs to operate these models. That will be a huge shock to the system.”