Knowledge Still Matters

The value of knowledge was never just in having the answer. It was in knowing what to do with it.

Last week I spent an afternoon at Swinburne University at a session on humanoid robots and the traction they’re starting to get, run with the ARC Research Hub for Humanoids. It was a genuinely interesting discussion, and the capabilities on show are further along than many people realise.

One conversation stayed with me afterwards: the challenge of replicating the human hand.

It’s not simply about getting fingers to move. It’s the dexterity, the sensitivity and the constant feedback that lets us pick up an egg without crushing it, or grip a hammer firmly enough to actually use it. We do most of this without consciously thinking about it, which is exactly what makes it so hard to reproduce in a machine.

After a couple of decades of watching technology evolve, that stuck with me. We tend to fixate on what a machine can do in a controlled demonstration — and it’s often the simple things, the bits we take for granted, that show how much is really going on underneath.

I’ve seen the same pattern in software and enterprise technology. A solution can look impressive, perform exactly as designed, and still fail to solve the problem a business actually has. Sometimes the problem was never properly understood in the first place.

AI brings this into sharper focus, because it’s now producing work that used to require considerable training and experience. Writing, coding, analysis and design are all changing. It’s fair to ask what happens to human expertise when machines can do so much of what we once relied on people for.

But I think we risk confusing the ability to produce an answer with the knowledge needed to use it well.

I come back to this whenever I see AI applied to a complex business problem. It can generate a design, analyse a market or draft a strategy in very little time, and that can be genuinely useful. Yet someone still has to decide whether the assumptions hold, whether the solution fits the operating environment, and what happens if it fails.

I’ve seen technically impressive solutions that simply weren’t right for the business. More capable tools don’t make that judgement any less important. If anything they raise the stakes, because we can now produce plausible answers much faster than we can properly evaluate them.

And when a decision has real consequences, somebody has to own it. That responsibility doesn’t disappear because the recommendation came from a machine.

One thing I’ve come to value in experienced people is their instinct for when a question needs challenging. They don’t necessarily know everything, but they’ve seen enough to know the information in front of them may not tell the whole story — and that instinct is usually earned the hard way, through decisions that worked, decisions that didn’t, and situations no textbook quite captured.

Experience can also make people stubborn. I’ve seen that too. Being experienced doesn’t automatically make you right, and there’s plenty AI can help us see that we’d otherwise miss. The opportunity is to put what the technology does well alongside the judgement people have built up, rather than assuming one replaces the other. There’s a related challenge for businesses that I don’t think gets enough attention.

A surprising amount of organisational knowledge sits with people who have been doing a job for years. They understand the exceptions, the history behind a decision, the reasons certain processes work the way they do. When they leave, the company often finds the documented procedure only told part of the story.

AI could be genuinely useful here. It gives us ways to capture knowledge, make information easier to find, and help less experienced people get answers without always having to track down the one person who knows everything.

But loading years of documents into a system doesn’t automatically transfer the experience behind them. Someone still needs to recognise when an answer is incomplete, when the circumstances are different, when the usual approach isn’t going to work.

There’s another question worth sitting with as we automate more work. How do people develop that experience in the first place?

If younger employees no longer do the routine work that previous generations learned their trade through, where does their practical understanding come from? Strip out the entry-level work today and you may be quietly removing the very thing that develops your senior people tomorrow.

I’m not arguing we should protect work that machines can do better, or slow down automation for its own sake. The productivity gains are real. But businesses need to look past the immediate cost saving and think about how they build expertise over time. After years of working with emerging technology, I’m comfortable saying machines will outperform people at a growing list of tasks. There’s little point pretending otherwise.

What I’m less convinced by is the idea that this makes human knowledge obsolete. We may come to place less value on remembering information or producing a standard answer. Understanding context, asking the right questions, and recognising when an answer doesn’t fit are a different thing altogether.

I left the campus thinking about the hand, but the point runs well beyond robotics. We consistently underestimate the knowledge behind the things people do almost unconsciously, whether that’s handling a physical object, diagnosing a fault, or making a judgement call after years in a particular field.

As the technology advances, we’ll need to get better at working out which parts of human expertise can be replicated, which can be enhanced, and which we risk losing without even noticing.

Knowledge still matters. The real challenge is recognising where its value actually lies, and making sure we don’t lose it while we’re busy automating everything else.