Artificial Intelligence (AI) will soon perform the essential tasks of software, the experts.
Sarah Friar, Chief Financial Officer for OpenAI, exclaimed the emerging role of AI-A-A-AS software engineer at a recent Goldman Sachs conference. OpenAi’s running AI agent, called A-Swe (Agentic Software Engineer), “is not only the current software engineers in your workforce, but instead is literally an agentic software engineer who can build an app for you. It can take a (pull application) that you can give to another engineer and build it.”
Also: why the new AI agent -tools from OpenAi can change how to cod
A-Swe not only builds the app, but “it does all the things that software engineers hate to do, it does its own quality assurance, its own bugtests and bug bashing and documentation,” Friar continued. “Things that you could never have Softwareers done. So suddenly you can suddenly multiply your workforce of your software engineering.”
Should software developers and engineers with tools such as A-Swe worry about their career perspectives? The reactions of the industry observers to the A-Swe-initiative Spectrum, from monitors pessimism to pragmatism.
Also: the best AI for coding in 2025 (and what cannot be used)
Software professionals “should be terrified,” Andy Thurai, technology strategist, former analyst at Constellation Research and former strategist at IBM Watson, told ZDNet. “The good will survive. The bad will have disappeared.”
Generative AI (Gen AI) “no longer only helps software developers and engineers; it redefines the nature of software development,” it once was Lori Schafer, CEO at Digital Wave. “In the next five years, IT organizations will see a dramatic shift from teams of developers who write code Line-by-Line to slimmer, more strategic teams of architects who orchestrate AI-generated programs.”
What this trend means is not necessarily wholesalers, but a major shift in the roles and priorities of software professionals. “With AI agents who produce fewer syntax errors, cleaner structure and faster iterations, software developers and engineers become editors and reviewers, not authors of every line,” Schafer said.
The rise of agentic AI in software “will probably not necessarily threaten job security immediately, but if you don’t know how to use AI agents, you might be threatened,” Thurai noted. “Think about this: one person does this entire app in less than a day, and the other takes four weeks to do the same. Who will survive longer? This trend also means fewer developers and software being hired.”
Also: AI agents are not only assistants: how they change the future of work today
Others suggested that AI agents will increase the skills for software development instead of replacing. OpenAIs A-Swe “represents an important progress in software development, but claiming that they can completely replace software engineers, is an exaggeration,” said Neil Sahota, CEO of ACSILABS and AI advisor at the United Nations.
“Although A-Swwe Code can write, it does not understand the ‘Why’ behind it. AI can simulate logic. However, it does not understand context, business nuance or edge cases that Real-World systems need. Generating (making, reading, updating and removing) is great, but it is a different approach to architect, safe restrictions.”
In large-scale companies or high bets, such as security, finance, health care and compliance, “we will still have human software engineers for a long time,” said Cassie Kozyrkov, CEO of Kozyr and former Chief Decision Scientist and data scientist at Google.
Software Engineering “requires more than just the rough opportunity to understand and write code,” says John Callery-Coyne, co-founder and main product and technology officer at Reflexai. “When AI companies perform these model benchmarks, they usually work in a vacuum, but real-life software engineering does not happen in a silo.”
Also: If AI agents multiply, it will be the new HR department
Effective software development requires “deep cooperation with other stakeholders, including researchers, designers and product managers, all of whom give input, often in real-time,” said Callery-Colyne. “Dialogues around nuanced product and user information will take place, and that context must be infused to make better code, which AI simply cannot do.”
The area where AIS and agents have been successful so far, “is that they do not work directly with customers, but instead help the most expensive part of IT, the programmers and software,” noted Thurai.
“Although 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 greatly shortens the coding time. This applies in particular for junior to developers at the middle level.”
AI software agents can be the most useful “when developers race during a major incident by time, to quickly roll out a fixed code and to put the systems back in use,” Thurai added. “But if the code is used in production as it is, this contributes to technical debts and the situation can ultimately make worse over the years, many incidents later.”
Also: will synthetic data derail the momentum of generative AI or are the breakthrough we need?
In addition, the new roles of software professionals must be investigated in an era of AI and agents. “Where performance matters, it is unlikely that software – engineering agents will eliminate the work – they will simply shift from writing the code to explaining and assessing it, which is not always a victory,” said Kozyrkov.
It is likely that software professionals “will play archaeologist in the mistakes of the AI,” Kozyrkov added. “Most coders will tell you that it is much nicer and fulfilling to write code yourself than reading someone else. Ai-generated work on a scale sounds great on paper, but someone will still have to check the bots, repair their mistakes, can evaluate margins, retain long-term systems and ultimately take responsibility.
