Scaling correctly comes first.

Scaling a business starts long before anyone opens a conversation about technology, and the order of those two things is most of the game. I have watched companies reach for automation the way someone reaches for a painkiller, hoping it will quiet a problem they have not actually named. It rarely does.

AI can make a process faster, leaner, and more consistent, but only once you have decided, in plain terms, what that process is meant to produce and what rules govern it. If the required output is still vague and the business logic beneath it is loose, automating that vagueness simply gives you the same confusion at greater speed. You do not get scale. You get a faster route to the same mess.

Part of the problem is that we use the word scale as though everyone means the same thing by it. We do not.

For some, scale means headcount: the comfort of a bigger team and a fuller office. For others, it means financial performance: the ability to do far more without the staff list growing in step. Those are not the same goal, and they do not require the same decisions.

Scale also looks different depending on the industry. A firm that sells hours and a firm that sells a product do not scale along the same curve, and pretending they do is how good intentions become expensive mistakes. Before any clever tooling enters the conversation, the first honest question is simply this: What does scale mean for this specific business, and how would we know when we have achieved it?

The deeper issue, the one I keep coming back to, is structural. A business reaches a scaling wall when the team, at every level, is the system.

When knowledge lives in people's heads, when the process depends on what a particular person happens to do on a particular day, and when nothing works unless the right individual is in the room, then the system is made of people. And a system made entirely of people can only grow in one direction: by adding more people.

You cannot make that kind of system meaningfully more efficient because there is nothing there to optimise except the humans themselves, and humans do not scale. To understand scale is to understand that the system itself must be scalable, designed so that output can grow without headcount increasing one-for-one. That is the real work. It is unglamorous, it does not photograph well, and at this point it has nothing to do with AI.

This is why I am wary of treating AI as the starting point. AI amplifies whatever system you already have.

Point it at a clear, well-defined, scalable process and it can extend your reach in ways that genuinely matter. Point it at a process that exists only in someone's habits and memory, and it will amplify the gaps just as faithfully. The framework for scale must come first, because the tool inherits the quality of the thing it is pointed at. 

I want to be fair to the optimism surrounding AI because it is not baseless. The efficiency gains are real and well documented.

Research from the University of St Andrews, based on a large survey of UK small and medium-sized businesses, found productivity gains from AI adoption ranging from roughly 27% to as much as 133%, with the largest improvements appearing, tellingly, among firms that started with the lowest productivity.

The upside is not marketing hype. It is measurable.

But notice what that finding is actually saying. The biggest gains go to businesses that had the most inefficiency to remove. That is a very different claim from the one often made in boardrooms, where AI is presented as a universal lever that lifts everyone equally. It does not. It rewards businesses that already understand where their processes are weak.

And this is where adoption and integration quietly part ways.

A recent Goldman Sachs survey of small business owners found that more than three-quarters were already using AI in some form, yet only 14% had genuinely embedded it into their core operations. Read that gap slowly because it captures most of the argument in a single statistic.

Almost everyone has adopted something. Almost no one has integrated it.

The distance between those two numbers is exactly the distance between buying the tool and doing the structural work required to make the tool matter. Adopting AI because everyone else is adopting AI, or because it feels embarrassing not to, delivers very little. It is motion mistaken for progress.

There is one belief in particular that I do not share, and it happens to be the loud one at the moment: that AI is here to replace people.

I think that framing is both wrong and unnecessarily narrow.

The same Goldman Sachs research found that around 87% of small businesses see AI as augmenting their workforce rather than displacing it, which matches what I see on the ground. The useful question was never, How many people can I remove?

It is this: If I have an employee whose contribution is worth R500,000, how do I use AI to help that person deliver the value of a R1 million employee? That is the mindset that compounds.

AI should be the thing that pushes a person, and a company, to grow into a more capable version of itself, not the thing that quietly thins the room. Replacement is a once-off saving. Elevation is a capability you keep.

None of this means AI is always the answer, and I would be dishonest to suggest otherwise. Optimisation has proven valuable across many areas of business, but there are industries and functions where AI is simply not the right route to scale.

In those environments, value sits in judgement, relationships, creativity, or craftsmanship that cannot be neatly compressed into a model.

Recognising those cases is not a fear of technology. It is the same discipline that underpins everything else discussed here: refusing to adopt a tool for its own sake and insisting that it earns its place against a clearly defined outcome.

So I will leave it where the thought that started this leaves it, because it remains the truest line I have on the subject.

You can chase every capability the technology industry is currently excited about, and there is a great deal worth being excited about. But the one thing that must be settled first is whether you can say, with real confidence rather than hope, that you are scaling correctly.

That answer comes from the structure of your business, from how clearly you have defined what it produces, and from how little of its success depends on a single irreplaceable person.

AI will sharpen that answer once you have it. It will not give it to you. That part is still ours to figure out.