The AI Shift: Preparing for What’s Next
AI tools can make developers faster — but at what cost to quality and creativity?
AI isn’t just writing code, it’s reshaping developer workflows. Fabrice Bellingard, VP of Product @ Sonar, reveals how to harness the good, mitigate the bad, and avoid the ugly pitfalls of AI-generated code.
We’re at an inflection point with AI. Already, this tech is revolutionizing how developers work as coding assistants gain popularity, helping developers work faster and accomplish more. With most developers already overwhelmed and experiencing nearly universal rates of burnout, AI can be seen as a welcomed relief. This is crucial, as business demand for software will only continue to increase: the market is projected to reach over $1586 billion by 2032.
But there’s a catch. AI isn’t a substitute for the human ability to think critically and creatively — it is complementary. Developers can use these AI tools to lighten their workloads, and they should. But it’s vital to remember that code reviews and analysis are still a critical part of the development process, and should be held to the same level of scrutiny and analysis as you would if written by a human. Perhaps even more so with AI, as it can’t be assumed that the output is reliable and secure as is.
Developers, though, aren’t and can’t be copy editors. To assume that the bulk of their time should go toward reviewing what AI writes isn’t the best way to use their expertise, and it devalues the necessary work they do for businesses. Leaders and teams instead need to look at this paradigm shift clear-eyed and find safe, responsible ways to integrate AI into existing workflows that will bolster code assurance. Development teams need to find the middle ground where AI can be used as an aid without sacrificing both quality and time on the job.
AI Isn’t Going Anywhere
Like it or not, AI is here to stay. Developers, and the businesses that hire them, can’t refuse to take advantage of AI in the software development process at a time. Its proliferation is unavoidable.
That said, AI does have inherent risks. These tools are helpful, but they can also make mistakes. Research indicates that about 50% of code generated by AI assistants, for instance, is incorrect or partially incorrect. That’s just one example of where AI has raised concern.
Developers must consider this and understand that AI-generated code cannot be taken at face value. That means implementing the proper guardrails to ensure that the AI code they commit and the software they release are secure and high-quality. Teams should follow a protocol of testing, analyzing, and conducting thorough code reviews before considering their software ready for deployment.
Testing is essential, and developers need time to do it — but this doesn’t mean they should turn into copy editors. Developers need to be able to focus their creativity and critical thinking on priority, revenue-generating software and coding projects. Right now, with the quality of AI-written code in question, they’re losing hours of that priority work to ensure the work AI does is up to par.
Reworking the Developer’s Role
Workflows should be augmented so that developers can focus on important projects that drive business results without being bogged down by code review or neglecting their priority work. Developers are hired for their expertise, which AI cannot yet replicate. So, what does it say about our software if we don’t focus on that fact?
The answer to this problem is automated tools. Automated tools that start in the integrated development environment (IDE) and help review code throughout the continuous development (CI/CD) pipeline will be pivotal in ensuring developers can do the jobs they were hired for while stopping bad code from reaching production. A “start left” approach — taking “shift left” a step further — using tools developers trust from the get-go, can help simplify the code examination process while reducing the time developers spend on grunt work.
Talent acquisition is another solution. Examining and potentially changing hiring processes and practices can help ensure developers aren’t so overwhelmed that they burn out entirely. Businesses must investigate their talent development programs and consider hiring outside of a group with the traditional, standard experience. They can evaluate developers from different spaces or those with non-traditional backgrounds and different mindsets, which can be extremely useful on the job.
The IT talent shortage is real. 56 percent of global leaders understand that and consider it a concern, especially as demand is expected to double by 2030. With less than half of schools offering foundational computer science education, the time to act is now. Developers’ jobs are changing due to the widespread use of AI tools, and hiring needs to change too. Combining updated hiring practices with giving developers the right tools is a solid path to offering burnout relief and supporting a better developer experience.
Dirty Code Runs, But That’s Not the Point
AI-generated code has another problem, too. It works initially but not further down the road when it matters most. Eventually, it breaks down, which is an expensive and reputationally damaging event. While more developers need to be trained and given the right tools to mitigate problems caused by AI-generated code, they must also shift their mindsets about what they feel comfortable deploying.
A Clean as You Code approach, in which developers constantly check the quality of their code as they write it, can help. That kind of approach is critical in ensuring businesses release software that runs and maintains quality long-term, especially since developers often rely on previously written code to help them speed along their work and meet business demands. They can’t afford to work with bad code that breaks down later.
Keeping a closer eye on code assurance and quality also impacts business. When poor-quality code is released, reputation is on the line. Dirty code is much more susceptible to breakdowns and security vulnerabilities. These breakages and breaches come with a cost: CISQ estimates the damage of poor software quality to be worth at least $2.4 trillion. It’s an expense no business can afford to weather, and one that can be avoided by paying more attention to code quality.
Our increased reliance on and transformation of AI technology — and coding assistant tools — comes with a catch. It requires a “trust and verify” approach to mitigate its many potential and significant risks. We can use AI coding assistants to help improve the developer experience, but we can’t rely on the output at face value. With proper tools in place to check the quality and security of AI-generated code, we can support innovation while maintaining the code quality essential to the software we depend on every day.
A dual approach of better tooling and redefining the developer’s role, workflow, and priorities will save businesses from the problem of faulty, dirty code and propel them into the AI-focused future. Developer experience will improve, and they will be more empowered to deliver the secure, reliable, and useful software needed for business growth for years to come.
Measuring Impact: The Gen AI Code Report (sponsored)
AI is shaking up software development, no question about it. But how do you know if it’s actually helping? Sure, AI can churn out code at lightning speed, but what about the hidden costs — like debugging bad code or wasting time on fixes?
That’s where LinearB’s Gen AI Code Report comes in. It’s your guide to making AI work for you, not against you.
See where AI fits in your SDLC
Measure real productivity gains (or losses)
Spot the risks before they become your problem
Don’t let AI be a wildcard in your workflow. Download the report and discover how to balance speed with quality while staying ahead of the game.
READ: The Gen AI Code Report