By Stefan Lauber, CEO, iFundi

Ask any engineering manager what keeps them up at night, and the answer hasn't changed in years: where is the next generation of engineers coming from? The Engineering Council of South Africa and the Department of Higher Education put the national shortfall at more than 40,000 qualified engineers, with critical gaps across every major discipline. Senior engineers are retiring and leaving the country faster than young ones can step up to replace them. Nobody running a mining, industrial, or infrastructure operation needs to be told this.

Now add AI. Handled badly, it's going to make the pipeline problem worse. Handled well, it might just help us fix it.

The real challenge is that people talk about AI as if it's just one thing getting steadily scarier. It is actually three different capabilities arriving on three different timelines and mixing them up only muddies the waters.

The first stage is the one every engineer already uses without thinking twice: a chat tool that drafts a paragraph or answers a quick question. That's already in the rearview mirror.

The second stage is arriving right now, and it's moving faster than almost anything I’ve seen in my career. Agentic AI doesn't just answer a question, it carries out multi-step tasks on its own. This isn't some futuristic theory for engineers anymore. Engineering software is already running AI agents that plan and execute workloads. In fact, Gartner projects that 40% of enterprise software will carry this kind of agent by year-end, up from under 5% just a year ago.

The third stage is physical, and it's already happening right here in South African mines. Anglo American's Kumba Iron Ore is using autonomous drilling and automatic truck braking at Sishen and Kolomela in the Northern Cape. That’s the true direction of travel: AI controlling physical equipment directly, not just producing documents. It’s on a longer timeline than the software layer, but it's coming.

Here's the truth: AI won't necessarily replace an engineer. But it will fundamentally change how engineers work.

Back in 1997, IBM's Deep Blue beat Garry Kasparov at chess, and everyone assumed computers had simply become better at the game than humans. What happened next was far more interesting: teams pairing an ordinary human player with a computer started beating both grandmasters and supercomputers working completely alone. The human's job wasn't to out-calculate the machine it was to direct it, deciding which suggestions to trust, when to look closer, and when to overrule it.

Every engineering manager should pay attention to this, because the same logic applies to your team. AI can do the heavy lifting on a first-pass calculation or an initial design right now. What a human engineer brings to the table is judgment knowing which results to trust, catching the mistake the model missed, and taking ultimate accountability for the work.

So, how do you develop engineering judgment? You get it the way engineers always have years of doing the real work, getting some of it wrong, and having a senior engineer catch the mistake and explain why. Junior drafts, senior reviews and corrects. That’s how a junior engineer transforms into someone capable of signing off on a safety-critical decision ten years down the line.

Crucially, that entire learning process runs on the exact routine technical work that AI is now best at handling. This isn't a new revelation. Back in the 1960s, economist Kenneth Arrow argued that people accumulate the judgment that makes them productive later precisely by doing repetitive tasks. Cut those tasks out, and you don’t just save a bit on a training budget you effectively cut the curriculum.

So if a company uses AI to bypass junior work rather than having them work alongside AI with a senior engineer still checking the output the cost goes well beyond a few entry-level headcount savings. It quietly switches off the only mechanism we have for producing engineers with sound judgment. Put that side-by-side with a national shortfall of 40,000 engineers, and it stops being a minor efficiency decision. It becomes a threat to whether we’ll have the talent we need in a decade.

None of this is an argument against using AI. It’s an argument against using it carelessly. According to research from McKinsey, top-performing companies succeed by keeping technology leaders closely involved in business strategy. They tie AI training directly to real projects, make sure engineering leadership and IT plan rollouts together, and keep junior engineers hands-on with senior oversight.

South Africa has spent years trying to fix its technical pipeline through structured, hands-on training. AI gives us the strongest reason yet to protect that work, not abandon it.