Threading the Needle on GenAI and Developer Experience
Generative AI can relieve developers of routine tasks, allowing them to concentrate on creative problem-solving, but it will not eliminate the need for skilled developers.
July 12, 2024
![robot typing on a laptop robot typing on a laptop](https://eu-images.contentstack.com/v3/assets/blt07f68461ccd75245/bltf467f079489fa278/66914f0596a129aa40e692b3/genAI-robot-typing-1716x965.jpg?width=1280&auto=webp&quality=95&format=jpg&disable=upscale)
Let's face it: A lot of what developers do is routine. I'm pretty sure I've never heard a developer declare, "Hooray, I get to set up a SQL database!" (Well … maybe I've heard that once or twice — I know a lot of developers who are connoisseurs of sarcasm.)
However, when it comes to the trivia, the scaffolding, the mandatory "setting of values that should be defaults but this library was written by idiots" … developers hate that. They'd much rather solve interesting problems than bash their heads against a framework that's broken but "the only one approved by legal."
TuringBots to the Rescue
The promise of TuringBots (generative AI [genAI] that improves the software development lifecycle) is that they will take developers away from the trivia, letting them focus on the knotty problems that require creative solutions. So far, we've seen that with code completion and code explanation. We're about to see genAI solutions provide help with much more: determining what to work on next, identifying problems in design, and even helping project managers understand when features will ship.
Most Enterprises Are Still Figuring Things Out
Organizations aren't sure they're seeing ROI yet. A lot weren't measuring developer productivity beforehand, so they can't tell if it has changed. Most academic studies have measured productivity with toy problems ("build an http server") or perceived productivity.
Many developers like having a genAI assist, but there's not a ton of proof that it makes things go faster. Confounding that: Developers quickly see life with genAI as the new normal. Story points get adjusted and new deadlines are set with the expectation of genAI help.
Introducing genAI into a development organization can be a challenge. Moving from deterministic development to probabilistic development needs training — more than the half-hour course your company's already done that says, "Don't put corporate data into public large language models." Senior developers struggle to change long-established methods of working. Junior developers need more guidance: They're taking on the role of code reviewer/debugger without having written much.
Some Organizations Are Over-Confident — and Will Pay the Price
Apart from the companies thinking about developer experience, some businesses are hoping they'll soon be able to replace their costly developers with developer robots. I'm certain that some will. When you see that, run. They're heading for cost overruns, maintenance nightmares, governance breakdowns, and vulnerabilities galore. GenAI makes it easy to create a lot of code that works most of the time. However, you still need experts to weed out the hallucinations.
Read "Improve Developer Experience With Generative AI" to understand how AI will help — but not replace — your developers.
— Andrew Cornwall, Senior Analyst, and Caroline Bonde, Senior Research Associate
About the Author
You May Also Like