AI Companies Want to Replace Software Engineers. What Does That Actually Mean?
By Andrew Lau , CEO, Jellyfish
Over the last year, we’ve seen AI companies and industry observers come out with more and more ambitious claims about what their technology can accomplish.
That’s right. The most recent moment of AI hype saw Cognition AI announce Devin, “the first AI software engineer.” While Devin has already been thoroughly debunked, we should expect the hype cycle to continue with bigger and bigger claims about what AI can do and who it will replace. No one is disputing that AI — not just generative coding tools like Copilot, but a range of targeted tools — is making an impact. As of last fall, Gartner found that nearly two-thirds (63%) of the software engineering leaders they surveyed had already adopted genAI tools as part of their tech stack, with another 30% saying they plan to in the future. In Jellyfish’s forthcoming 2024 State of Engineering Management, 52% engineers say their team has embraced AI; 79% of those who have embraced AI are using generative AI specifically. The technology is gaining ground.
But can AI ever fully replace a software engineer? It begs the question: what is a software engineer? The answer isn’t as straightforward as an AI sales pitch would have you believe.
What is a software engineer?
Even more than asking what a software engineer is, it might be more useful to ask what a software engineer does. Is the software engineer’s role just to write the code? The AI tools we’re talking about here — even the ones that claim to replace the engineer entirely — deal with code generation. But does a software engineer’s job begin and end with generating code?
Of course not.
A software engineer is part of a larger organization, and their role includes a lot of responsibilities that take place both before and after the actual code is written. The engineer will collect feedback from the product manager on where the software needs to grow and change. The engineer will make predictions on when their work will be finished and provide updates to their team leaders to make sure the rest of the organization reacts accordingly. The engineer is responsible for a whole lot of communication and teamwork that isn’t so easily replaced by genAI.
Is it possible that all of these tasks will one day be automated away? Maybe, but I’m pretty skeptical.
领英推荐
The work of an engineering team isn’t so predictable or repetitive that it can simply be replaced by an AI module. There’s real creativity involved: deep understanding of what the customer needs and how the product can best be built to solve it. These teams are also responsible for brokering trade-offs on a daily basis and ensuring that engineering work is aligned with the overall direction of the business. An engineer’s role isn’t simply to take orders and write code — they’re involved in two-way communication and team-level planning that cannot be replaced by AI.
While engineering organizations today are agonizing over the question of whether they’ll have to replace human engineers with generative AI, I’d suggest that’s the wrong way of looking at the situation. Instead of asking yourself whether and when you’ll need to replace engineers, start trying to identify where the line is going to be drawn between AI and human engineering, and what tasks will be automated away for good.
Where’s the line and what does it mean?
It’s still early days for genAI, but it’s time for both teams and individuals to move past broad experiments and start making some specific bets around how AI is going to change their operations. Where can engineers get ahead of the game by taking advantage of AI, and where do they need to buttress their skills to stay relevant in the future? Let’s make some guesses.?
We know genAI is good at generating — it can create code quickly, but can it create the right code in the eyes of the product, the business, and the customer? As companies adopt different AI modules, apps, and even products, we don’t know to what extent these can deliver value directly to the end user and to what extent they require intervention. These tools are prone to hallucinating, and the jury is still out on what genAI means for code quality. Skill sets need to shift as a result. Engineers are going to be spending less time coding and more time reviewing. They’ll be paying less attention to the day-to-day generation of code and much more time thinking about architecture and about the product as a whole. For individual engineers, the goal should be to become indispensable either at the fine details — reviewing and perfecting AI-generated code — or at the big picture. Everything in between is going to be disrupted.
For engineering leaders and employers, it’s time to start predicting how your team is going to evolve. If you know you need more people that can think about the product at the architecture level, how should your interviews change accordingly? What backgrounds are going to prove most valuable in the next five to ten years? A strong education in computer science is probably going to be much more valuable now than your standard bootcamp experience. Finding ways to identify creative thinking will pay huge dividends as more repetitive work is automated away.
Once you start getting into the fine details of making predictions and plans, none of this is as scary as it seems. The more prepared you are for how this technology could evolve in the future, the more likely you’ll be to outperform your competition.
What will genAI be worth to engineering teams?
There’s one other factor we can’t afford to ignore when it comes to long-term planning and genAI. What is this all going to cost?
We still don’t know the extent to which engineering teams will be able to rely on genAI in their day-to-day work, and we don’t have a clear sense of what the costs are going to look like for a small- or mid-sized engineering team. Is genAI going to be something that’s accessible to every engineer on the planet, or is it going to belong to only the largest, most well resourced organizations?
As companies start bringing genAI products to market and start putting a price on their offerings, we’ll see customers ask themselves how much these services are actually worth. Every engineering team will need to know not only what their software engineers do but also how much they cost. That’s where we’ll truly find the line between where AI ends and where the real work of being a software engineer begins.