A few days ago, during a Data Leaders call, I found myself thinking about a quote I’d first seen years ago and that someone brought up in the discussion. It came from Harvard Business Review in 2017, long before ChatGPT, foundation models or AI copilots had entered the mainstream.
“If your company isn’t good at analytics, it’s not ready for AI.”
At the time, it seemed like a perfectly sensible observation. Analytics was the stepping stone towards machine learning, and machine learning was still largely the preserve of data scientists rather than the concern of every executive team.
Reading it again today, though, it strikes me as something rather more interesting. The technology has changed almost beyond recognition, yet the advice hasn’t aged at all. If anything, it has become more relevant.
That feels counterintuitive. We are living through what is probably the fastest period of technological development many of us have experienced. Every week brings another model with better reasoning, a lower price (or a different pricing model), or some remarkable new capability. The conversation has become one of speed: who is deploying AI, who is falling behind, and how quickly organisations can get something into production. We have actually seen this before, for example when Data Science was the “Sexiest Job of the 21st Century”, with everyone trying to implement data science workflows, because everyone else was doing it.
And so, I’ve started to wonder whether we’re measuring the wrong thing.
Whenever I speak to organisations about AI strategy, there is an almost inevitable point in the conversation where someone asks how quickly they can deploy AI. It is an entirely reasonable question, except that it assumes AI is the destination. I’m increasingly convinced it isn’t. AI is simply the latest occupant of a much taller stack, and everything underneath that stack still matters just as much as it did before. Perhaps even more.
One of the unintended consequences of the generative AI boom is that it has made intelligence look deceptively easy. You can open a browser, type a prompt, and receive something that appears thoughtful, articulate and often surprisingly useful. It’s natural to assume that if “”intelligence“” has become this accessible, then integrating it into an organisation should be equally straightforward. Unfortunately, or not actually, organisations are rather more complicated than chat windows.
Over the years I’ve worked with enough data projects to know that technology rarely fails because the algorithms are inadequate. More often than not, the problems are much more mundane. The data means different things in different systems, and to different people. Business processes have evolved independently across departments. Nobody quite agrees which metric should be trusted. A customer might exist five times under five slightly different names, and every report gives a slightly different answer because each team has quietly developed its own definition of the truth. None of those problems are new. Neither does AI solve them all in all cases. In fact, it has an awkward habit of exposing them.
That’s why I sometimes think we’ve misunderstood what AI actually does inside an organisation. We talk about it as though it creates competitive advantage, when in reality it often reveals whether competitive advantage already existed. Let me explain.
Imagine two organisations deploying exactly the same model. They have access to the same algorithms, the same computing infrastructure and, in many cases, the same underlying foundation model. Yet one delivers measurable business value within months while the other struggles through endless pilots that never quite reach production.
It is tempting to explain the difference in terms of AI expertise. I don’t think that’s the whole story. More often, the difference was established years before anyone mentioned large language models.
One organisation invested patiently in analytics. It improved data quality because accurate reporting mattered, not because AI demanded it. It standardised processes because operational efficiency was valuable in its own right. It developed engineering disciplines that made deploying software predictable and repeatable. Governance wasn’t introduced to satisfy regulators; it existed because well-run organisations need to know who owns decisions and why.
Then AI arrived.
To the outside world, it looked as though the organisation had moved exceptionally quickly. In reality, it had simply spent years preparing the ground.
The other organisation now finds itself trying to retrofit those same capabilities beneath an AI strategy that is already under way. That is a much harder problem. Retrofitting almost always is. Software engineers have known this for decades. Changing the foundations of a system after the building is standing is invariably more expensive than getting them right in the first place. Ask me about it!
I suspect we’ll see exactly the same pattern over the next few years. The companies celebrated today for launching AI assistants and autonomous agents may not necessarily be the ones that dominate tomorrow. The winners are more likely to be those that quietly invested in the less glamorous work that rarely appears in keynote presentations: data governance, analytics, architecture, engineering practices, organisational learning and clear operational processes.
None of those topics generate headlines. They may produce impressive demonstrations for shareholders. Yet they determine whether AI becomes a sustainable capability or merely an expensive experiment.
Perhaps that’s why the quote from 2017 has stayed with me for so long. At first glance it looks like a comment about analytics. I don’t think it is. It’s really a comment about organisational maturity.
The competitive advantage was never going to belong to those who adopted AI first. As these technologies become cheaper and more accessible, everyone will eventually have access to similar models. The advantage lies elsewhere.
It belongs to the organisations that spent years building foundations while everyone else was looking for shortcuts.
And perhaps that’s the irony of this particular technological revolution. We keep talking about artificial intelligence as though it changes everything. In many respects it does. But the organisations that will benefit the most are likely to discover that success depended on the same things it always did: good data, sound engineering, disciplined thinking, and a willingness to invest in capabilities long before they become fashionable.
AI hasn’t rewritten the rules.
It’s simply made them impossible to ignore.