AI use in software development is at an unusual stage of maturity. The overwhelming majority of developers are using AI tools to create code, but there are a lot of questions about the ramifications of doing so. How much should developers trust the code that AI produces? Where specifically does AI belong in the software development process? And what exactly is the role of software developers in the age of AI?
According to McKinsey’s The State of AI in 2026, 80% of AI users say that it has improved their individual productivity, and 50% say it helps them make better decisions. Yet there’s not a universally positive response from those who are making the most use of AI technology.
Some have pointed to a kind of AI “fatigue”, where developers managing multiple agents, juggling different tasks, and continually switching context are quickly overwhelmed.
Creating and maintaining a happy team is no small feat for any business. The way AI is changing the developer role makes it even trickier.
Changing roles
AI is often compared to a junior developer in its abilities. It follows, therefore, that those using AI are effectively managers, responsible for prompting the AI, checking in on progress, reviewing the output, and offering feedback where needed.
But most developers are not trained as managers, and even if they were, not all of the principles of management apply to AI agents. Some developers are finding the responsibility of managing multiple agents exhausting.
Part of this is from the sheer volume of code that AI is capable of producing. Running five agents at once will mean producing five times as much code. It seems very efficient, but it also means five times as much code that needs to be reviewed and approved.
This isn’t really a new problem. We have more automation and convenience in our lives than ever before, from washing machines to automatic bill payment to delivery apps, but few people claim that they have too much free time.
The problem is not AI or convenience, but is wider, and systemic. In our daily lives, automation allowed us to take on other responsibilities, taking up the free time we were promised. In development teams, part of the problem is that the infrastructure and processes around AI have adapted, and bottlenecks have just shifted around.
AI tools have been built for one-to-one conversations and individual productivity. The pattern of adoption has reinforced this, as individuals experiment with AI tools in their own tasks, siloed away from everyone else. But that’s not how most modern development teams work.
The solution is not to automate review, but to make AI agents less like personal assistants and more like part of the team: running in the cloud, triggered by events as well as user prompts, running in the background where possible, and assisting across the whole workflow, such as in review and validation. The human team can then focus more on what they are better equipped to do, rather than micromanaging every AI task.
The shift in role shouldn’t be switching one type of drudgery for another, creating and refining prompts over and over instead of manually coding. The shift will be to one of scoping, architecture, and designing the system, not necessarily the code.
Better measurement
How do organizations measure success? Metrics such as “lines of code” and “velocity” have sometimes been used but are often criticized as blunt and uninformative. Better frameworks have been developed, such as DORA, but these may also have to change as good metrics as AI-assisted development becomes the norm.
What these metrics may miss is how developers judge themselves. Previously, a developer could pride themselves on particularly elegant code or solving an issue through a combination of their coding experience and smart thinking. The use of AI blurs the lines of who did what.
If organizations can navigate the change so that AI is creating real efficiency, developers may still feel that they are merely corralling AI, which is doing the “real work”.
Adopting AI means a cultural change as well as developing a mature approach to integrating AI assistance into workflows. Developers will need to understand the full scope of their roles and responsibilities. They will need to do much more than “prompt engineering.”
They will need to fully scope out projects, provide the context, and collaborate with both their peers and AI agents. All of this is vital to a project’s success, but it requires a shift in mindset for those who are used to measuring their value in terms of the code they produce.
Maintaining morale and team happiness means a focus on company culture as much as technology adoption. Teams should examine how team members are recognized for the work they do and whether these metrics reflect changing roles.
Culture is important not just to make people feel valued, but to ensure that people don’t get overwhelmed by increasing amounts of work created by AI. This is especially true if they feel unable to share their frustration and are “siloed” and unable to look to others for assistance.
Successful AI adoption is about more than who is using it. It’s about how well it’s integrated, how that is affecting efficiency overall, and how those on the front line are dealing with the changes to their role. Morale and happiness are just as important of a measurement as any other.
We’ve featured the best laptop for programming.
This article was produced as part ofTechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.
The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here:https://www.techradar.com/pro/perspectives-how-to-submit
