The Future of System Design

AI’s usefulness isn’t in question – but is it a step change? Right now, probably not. Not yet, anyway. There’s no doubt it will get there, but like most shifts in technology, adoption and impact take time.

It reminds me of when the iPhone first landed. At launch, it was a clever fusion of your iPod (remember those?), your phone, and a basic camera. But the thing that made it revolutionary – the app ecosystem that’s now embedded in almost every part of our lives – didn’t exist back then. That came later, as developers and users started pushing the limits. Now we look back and wonder how we ever lived without it.

That’s where we are with AI. The foundations are in place, and what comes next could be transformative. In a decade’s time, we’ll likely look back and ask: how did we run our businesses without this?

On a related note, I was talking to a client in the client lifecycle space, and something stood out. Most systems today are still designed entirely around human users. UX remains a major focus – and rightly so – because when we design for people, we tend to get better results.

But what if not all users are people?

Now we’re entertaining a phase where digital agents can handle specific parts of workflows. Not just automating repetitive tasks – but engaging with customers in natural language, making decisions, acting independently in the system.

That raises an important question: is this really new, or just automation rebranded?

The key difference lies in how we design. Simply adding GenAI into existing systems might offer short-term gains, but it doesn’t unlock the full value – and in some cases, it can backfire. You risk applying a probabilistic model to a task that demands a precise, deterministic outcome. That’s a real problem, especially in regulated environments where there’s no room for “probably correct”.

If we designed systems from the ground up to include both human and digital agents, we’d make different decisions. We’d think more critically about control frameworks, about how different AI models interact, and about how to ensure accountability when AI becomes part of the customer interface.

And then there’s brand. If you’re letting digital agents speak to customers on your behalf, the tone, consistency, and experience have to reflect the brand you’ve spent years – and likely millions – building. It only takes one off-key interaction to undo a lot of that hard work.

The more I explore this, the more convinced I am that systems will look fundamentally different when they’re designed with automation at the core. The challenge, as with most innovations, is that we’re anchored by what we already know. When the iPhone came out, did anyone really expect it would replace your wallet? (Well… everywhere but the US, weird that).

So yes – AI is moving fast, and the potential is exciting. But to really take advantage of what it offers, we need to go beyond plugging it in. We need to rethink how we design the system that will depend on it.