This post was originally published on this site.
Andrew Lau is cofounder and CEO of Jellyfish, a pioneering engineering management platform.
“Copilot writes all of my code now.”
“AI is coming for our jobs.”
“We’re not going to need software engineers in two years.”
Engineers, relax. The AI coding takeover isn’t here—yet.
Yes, we’re seeing widespread adoption of AI coding tools. Three out of every four developers are using or plan to use AI tools in their engineering work. Half of the Fortune 500 companies are Copilot customers. These tools are undoubtedly gaining traction, particularly as companies start to measure their ROI.
But just because AI coding tools are delivering value and growing in popularity doesn’t mean engineers are on the verge of a mass extinction event. Here are three reasons why.
1. Most organizations are just getting started
Major technology shifts don’t happen overnight. In particular, large organizations move on steady, deliberate timelines. Take this example from a recent McKinsey article: An automotive organization rolled out an AI coding tool to 10,000 developers, later to find that just one-fifth of their engineering organization actually used the tool.
Many companies are still in the early, uncertain stages of AI adoption, and it will likely be another year or so before we see organizations pushing for large-scale deployments. And we’re still years away from broad, market-wide adoption of AI coding assistants. Until then, we shouldn’t expect dramatic changes to engineer headcount.
2. AI coding tools are driving modest boosts in productivity.
Are engineers more effective when they use AI? Do the tools justify their price tag? The industry is still trying to figure this out. Based on anonymized data from nearly 15,000 Copilot users from our dashboard, we found modest increases in coding speed: Copilot users deliver code 12.6% faster than those who aren’t using AI. A 12.6% increase is great, but there’s still a long way to go.
We’re in a gray area. Yes, these tools seem magical and are certainly promising enough to justify continued investment, but they aren’t yet delivering results that would drive engineering leaders to make sweeping, permanent changes.
3. An engineer’s job is more than just coding.
If an engineer’s job consisted 100% of hands-on-keyboard coding, then we might have reason to be concerned. But the job is much more than just code generation: Based on my experience both as an engineer and as a leader of engineering teams, I’d estimate that engineers spend less than half their time on coding itself. An engineer’s work includes so many things that can’t be accomplished through generative AI alone: code reviews, architecture, stand-ups, interviews and brainstorms. AI can automate some of an engineer’s work, but you need humans to provide vision and align engineering work to the larger organization and its goals.
Of course, the market continues to evolve: New tools are cropping up to help in areas not currently supported by AI, and AI vendors will continue to innovate. Although it’s natural to worry about what AI means for engineers, our time will be better spent focusing on how to integrate these emerging tools to increase the value engineers already contribute to their organizations.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?