How is AI killing jobs – while software hiring is surging?

The software job market is sending a confusing message to employers.  

Amid continuing headlines about AI annihilating white-collar tech jobs, one U.S. analytics firm has tracked more than 67,000 current software engineering openings.  

Software engineer postings have roughly doubled since mid-2023 at tech companies, and have jumped about 30 percent this year so far, according to Business Insider. 

According to experts it’s a crucial moment for tech leaders to have a fundamental rethink of their recruiting and development of software talent. 

AI is changing what “entry level” means 

Generative AI is already stripping away much of the repetitive work that once defined junior developer roles; according to Qusay Mahmoud, professor and assistant dean of engineering at Ontario Tech University, the traditional “first job” in software – fixing small bugs, writing boilerplate code and doing basic testing – is no longer the same on-ramp it used to be. 

“I am hearing from industry partners that what’s disappearing are the routine tasks that used to define junior software development roles,” Mahmoud says. 

“AI can now handle a meaningful portion of that work.” 

For employers, that raises the bar for expectations in early career recruitment. Instead of looking for candidates who are simply willing to “pay their dues” on repetitive tasks or basic tasks, employers need people who can immediately contribute to design discussions, risk assessments and AI-assisted workflows. 

This means seeking very different – and wider – skill profiles at the start of the software career ladder, including understanding whole systems, analyzing output and being aware of security, reliability and privacy.  

“Employers now expect software engineers who can work beyond the code,” Mahmoud says. 

“The question is no longer whether someone can write code, but understanding systems, exercising judgment, and taking responsibility for part of them.” 

‘Deep engineers’ and cross-sector redesign 

These heightened expectations are showing up across sectors, not only in pure-play tech companies; semiconductor, healthcare and financial services employers are all experimenting with AI in development work, and discovering that entry-level roles need to be redesigned, not removed.  

Triparna Chakraborty, an engineer-turned HR business partner in the semiconductor industry, sees the same pattern inside large technical teams; employers are increasingly looking for “deep engineers with AI fluency,” she says, who can work across hardware, firmware and software and understand how AI workloads move through infrastructure, especially as companies build out data centres and AI infrastructure. 

“There is growing expectation across industries from talents, including early-in-career talents, to build using AI,” Chakraborty says.  

“For example, can they break down a problem into a solution framework that can also help other teams? Can they use AI tools as an accelerator? This is far more valuable.” 

AI also creates fresh regulatory, privacy and security questions that cannot be left to automated systems alone, and Chakraborty makes a distinction between tasks and roles, which she links directly to hiring decisions.  

“There is a growing notion that AI is replacing software engineers. But AI is replacing repetitive coding tasks,” she says. 

“What it is not replacing are tech talents and engineers that understand the problem behind that code. The ones who can use AI to accelerate their work while staying deeply connected to the customer problem.” 

Rewrite job descriptions around systems, not tools 

For HR and talent acquisition teams, one of the most immediate levers is the job posting itself. In many organizations, requisitions for developers still read like they did five or 10 years ago, with long lists of languages and frameworks and generic “years of experience” requirements.  

Mahmoud says many software requisitions still assume a pre-AI reality. 

“They overemphasize tools and years of experience,” he says. “Employers should focus on capabilities … instead of asking if someone knows a specific framework, they should ask the candidate whether they can build, evaluate, and improve a system under real constraints.” 

Chakraborty urges HR teams to strip back long lists of programming languages and platforms in favour of concrete problems to be solved.  

“Stop listing tools and languages. Tools are changing every six months. Instead, lead with a specific business case or problem,” she says. “A job description that says: ‘You will be expected to diagnose XYZ problems across ABC domains’ is more powerful than ‘know MATLAB, PYTHON’.” 

Hire for tomorrow, not today 

Chakraborty also emphasizes hiring explicitly for long-term scope; for organizations that expect to change cloud providers, AI platforms or front-end frameworks over the next few years, anchoring requirements to specific tools can quickly make postings obsolete.  

“Tools are evolving and job descriptions are changing rapidly, so look for engineers who have worked across multiple domains,” Chakraborty says.   

“Technical depth still matters, but ability to apply that depth across changing context matters even more. That separates a good hire from a great one.” 

Essentially, recruiters need to probe how candidates think, not just what they have built. As Mahmoud explains, screening for “real” experience versus AI-assisted output is basic stakes, and is less about catching people out and more about understanding how they approach complexity, trade-offs and fails. 

“The clearest signal is depth of explanation,” he says.  

“If a portfolio includes GitHub or similar platforms, that process may also be visible in the commits, documentation, and iteration. Strong candidates show how they built and improved the system, and not just the final result.” 

Manager training and involvement is the key 

Plus, organizations that benefit most will be the ones that pair disciplined hiring with internal investments into knowledge and training; Chakraborty says the most effective AI programs start with managers, not one-size-fits-all courses. 

Instead of large-scale AI suite adoptions and trainings, employers should work closely with managers on how tools can augment their own teams will be the most effective, she says. 

“Managers are most incentivized to use AI and scale it to their teams,” she explains. 

“So creating programs that open up platforms and resources for managers to work through 'what AI means for them' will be key.”