Artificial intelligence (AI) will soon be performing the essential tasks of software engineers — or so the experts say.
Sarah Friar, chief financial officer for OpenAI, proclaimed AI-as-software-engineer’s emerging role at a recent Goldman Sachs conference. OpenAI’s pending AI agent, called A-SWE (Agentic Software Engineer), “is not just augmenting the current software engineers in your workforce, but instead is literally an agentic software engineer that can build an app for you. It can take a [pull request] that you can give to any other engineer and go build it.”
Also: Why OpenAI’s new AI agent tools could change how you code
Not only does A-SWE build the app, but “it does all the things that software engineers hate to do, it does its own quality assurance, its own bug testing and bug bashing, and documentation,” Friar continued. “Things that you could never get software engineers to do. So suddenly you can force multiply your software engineering workforce.”
With tools such as A-SWE emerging, should software developers and engineers be worried about their career prospects? Industry observers’ reactions to the A-SWE initiative span the spectrum, from guarded pessimism to pragmatism.
Also: The best AI for coding in 2025 (and what not to use)
Software professionals “should be terrified,” Andy Thurai, technology strategist, former analyst with Constellation Research, and former strategist with IBM Watson, told ZDNET. “The good ones will survive. The bad ones will be gone.”
Generative AI (Gen AI) “no longer just assists software developers and engineers; it’s redefining the very nature of software development,” agreed Lori Schafer, CEO at Digital Wave. “In the next five years, IT organizations will see a dramatic shift from teams of developers writing code line-by-line to leaner, more strategic teams of architects orchestrating AI-generated programs.”
What this trend means isn’t necessarily wholesale job replacements, but a major shift in the roles and priorities of software professionals. “With AI agents producing fewer syntax errors, cleaner structure, and faster iterations, software developers and engineers are becoming editors and reviewers, not authors of every line,” Schafer said.
<!–>
The rise of agentic AI in software “probably won’t threaten job security per se immediately, but if you don’t know how to use AI agents, then you might be threatened,” Thurai pointed out. “Think of this: one person does this entire app in under a day, and the other takes four weeks to do the same thing. Who will survive longer? This trend will also mean fewer devs and software engineers will get hired.”
Also: AI agents aren’t just assistants: How they’re changing the future of work today
Others suggested AI agents will augment rather than replace software development skills. OpenAI’s A-SWE “represents a significant advancement in software development, but asserting that they can fully replace software engineers is an overstatement,” said Neil Sahota, CEO at ACSILabs and AI advisor to the United Nations.
“While A-SWE can write code, it doesn’t understand the ‘why’ behind it. AI can mimic logic. However, it doesn’t grasp context, business nuance, or edge cases that real-world systems need. Generating [create, read, update, and delete] operations is great, but it’s a different approach to architect scalable, secure solutions under tight constraints.”
In large-scale enterprises or high-stakes domains, such as security, finance, healthcare, and compliance, “we’ll have human software engineers in the loop for a long time to come,” said Cassie Kozyrkov, CEO of Kozyr and former chief decision scientist and data scientist at Google.
Software engineering “requires more than just the raw ability to understand and write code,” said John Callery-Coyne, co-founder and chief product and technology officer at ReflexAI. “When AI companies are running these model benchmarks, they’re usually working in a vacuum, but real-life software engineering doesn’t happen in a silo.”
Also: As AI agents multiply, IT becomes the new HR department
Effective software development requires “deep collaboration with other stakeholders, including researchers, designers, and product managers, who are all giving input, often in real time,” said Callery-Colyne. “Dialogues around nuanced product and user information will occur, and that context must be infused into creating better code, which is something AI simply cannot do.”
The area where AIs and agents have been successful so far, “is that they don’t work with customers directly, but instead assist the most expensive part of any IT, the programmers and software engineers,” Thurai pointed out.
“While the accuracy has improved over the years, Gen AI is still not 100% accurate. But based on my conversations with many enterprise developers, the technology cuts down coding time tremendously. This is especially true for junior to mid-senior level developers.”
AI software agents may be most helpful “when developers are racing against time during a major incident, to roll out a fixed code quickly, and have the systems back up and running,” Thurai added. “But if the code is deployed in production as is, then it adds to tech debt and could eventually make the situation worse over the years, many incidents later.”
Also: Will synthetic data derail generative AI’s momentum or be the breakthrough we need?
In addition, the new roles of software professionals in an age of AI and agents need to be explored. “Where performance matters, software engineering agents are unlikely to eliminate the work – they’ll just shift it from writing the code to explaining and reviewing it, which isn’t always a win,” Kozyrkov said.
It’s likely that software professionals “will find themselves playing archeologist in the AI’s mistakes,” Kozyrkov added. “Most coders will tell you it’s far more fun and fulfilling to write code yourself than read someone else’s. AI-generated labor at scale sounds great on paper, but someone will still need to monitor the bots, fix their mistakes, evaluate edge cases, maintain long-term systems, and ultimately take responsibility. Unless we’re careful, we risk replacing builders with babysitters. It’s up to us how that plays out.”
