The AI skills gap in engineering isn’t what most companies think it is

In June 2023, GitHub published research showing developers using Copilot completed tasks 55% faster than those working without it. Not “AI might speed things up eventually” — 55% faster, measurable, in a controlled study with 95 professional developers across real tasks. McKinsey’s own analysis followed: 45–50% of time saved across software engineering activities when generative AI tools are properly embedded. These aren’t projections. They’re already happening in teams right now.

Something changed in 2023 that a lot of engineering hiring hasn’t caught up with yet.

And yet the majority of engineering job descriptions I see still look roughly like they did in 2021. There’s occasionally a line about “experience with AI tools” somewhere near the bottom. That’s not a strategy. That’s a checkbox.

---

The problem isn’t awareness. It’s knowing what to actually hire for.

Most hiring managers I speak to have noticed AI is shifting things. What they’re less clear on is what “AI-ready” actually means in practice. There are broadly two failure modes.

The first is hiring for hype. Someone reads that prompt engineering is a skill, adds it to the job description, and hopes for the best. Or they decide they need a dedicated “AI engineer” without being clear what problem that person is actually solving. These hires often underdeliver because the brief wasn’t specific enough about the work.

The second is doing nothing. Assuming the team will figure it out. Assuming the tools will level out and the skill gap will close itself. The Stack Overflow 2024 Developer Survey found 62% of developers are already actively using AI tools in their work. The gap between teams who’ve properly embedded AI into their workflow and those who haven’t is growing. Waiting for it to resolve is not a plan.

---

What “AI-native” actually looks like in practice

It’s not someone who’s done a prompt engineering course. The engineers genuinely shifting what’s possible aren’t the ones with the most AI theory behind them. They’re the ones who’ve worked out where AI accelerates their existing process and where it introduces risk. They’re sceptical where they should be, experimental where it matters, and they’re usually already doing it without anyone asking them to.

In interviews, you’ll spot them by how they talk about the tools. Not as magic. Not with dismissal. Practically. “We used Copilot on the data transformation layer — saved about two weeks — but we reviewed every output because the edge cases were messy.” That’s different from “I’ve been experimenting with ChatGPT.”

---

The talent pool is smaller than the noise suggests

LinkedIn data shows AI skill mentions on profiles have grown 142 times over the past eight years. Sounds like there’s plenty of talent out there. In practice, a large proportion of that is self-reported exposure rather than embedded practice. The engineers who’ve actually changed how they work because of AI — not just added a buzzword to their profile — are a specific group and they tend to be employed.

A Boston Consulting Group study found workers applying AI to complex tasks performed 40% better than those who didn’t. The people responsible for those numbers in engineering teams know their value. They’re not applying to your job advert because the job advert doesn’t speak to them.

---

What this means practically for your next hire

The brief matters more than it ever did. “Software engineer, AI experience preferred” won’t reach the right people and won’t help you assess the ones who do apply. Being specific about your stack, where AI is already being used in your workflow, and what you’re actually trying to achieve — that changes who responds and what good looks like.

It also changes how you assess. Someone who’s spent the last two years working in an AI-augmented engineering environment is going to look different on paper from a senior engineer who hasn’t. The CV screening logic that worked fine in 2021 will screen out the person you actually want in 2025.

The World Economic Forum’s Future of Jobs report is pretty direct about it: 44% of workers’ core skills are expected to be disrupted within the next five years. In software, that disruption is already underway. It’s not coming.

Getting the brief right, knowing where to look, and assessing properly isn’t a small operational detail. For engineering teams right now, it’s the difference between building something AI-ready and just calling it that.

---

Sources: GitHub Copilot Research Study (2023) · McKinsey Global Institute — The Economic Potential of Generative AI (2023) · Stack Overflow Developer Survey (2024) · Boston Consulting Group — AI at Work (2023) · LinkedIn Talent Insights · World Economic Forum — Future of Jobs Report (2025)