Why Your AI Strategy Needs a Data Foundations Chapter
There's a version of the AI conversation happening in boardrooms across New Zealand right now that goes something like this: the vendor presents the capability, the numbers look compelling, the leadership team agrees it's worth pursuing, and someone is tasked with "getting AI moving." Six months later, the project is running behind, the business case is under pressure, and there's a growing sense that the technology just didn't live up to the hype.
Except it usually isn't the technology's fault.
The problem, almost every time, is that nobody stopped to ask what the AI would actually run on. And the answer, when you look honestly, is data that was never built to carry that load. Inconsistent definitions across systems. Pipelines that work until they don't. Fields that mean different things depending on who you ask. Data that people in the business are quietly suspicious of but have never had the mandate to fix.
AI doesn't smooth over any of that. It runs on it, directly. And if the foundation is shaky, the outputs will be too.
What "Data Foundations" Actually Means
It's worth being specific, because the phrase gets used loosely. Data foundations aren't about collecting more data or building more integrations. It's about what you can reliably trust. A strong data foundation means your key business concepts - customers, products, transactions, whatever matters most to your decisions - are defined clearly, stored consistently, and understood the same way by the people and systems that use them.
It means your data pipelines are monitored, so you know when something breaks before a decision is made on stale numbers. It means there's governance: someone owns each data asset, someone's responsible when quality slips, and there's a process for resolving the inevitable disagreements about what a metric actually means.
None of this is glamorous. But it's the difference between an AI system you can trust and one that generates confident-sounding answers you can't verify.
The Organisations Getting This Right
We work with NZ organisations across health, utilities, local government, and financial services. The ones who are furthest ahead with AI aren't necessarily the ones who moved first on a model or a platform. They're the ones who, somewhere in the last two or three years, made a deliberate decision to get their data house in order.
One client, a group operating across 30 independent businesses, spent several months standardising how financial data was defined and consolidated across the network. At the time, it felt like plumbing. Now, it's what gives their leadership team real-time visibility across the whole portfolio, and it's the foundation their next wave of AI-powered decision-making will sit on. That year of groundwork wasn't a cost of doing AI. It was the prerequisite for doing it well.
The Question Worth Asking Before Your Next AI Conversation
If you're heading into a board conversation about AI capability, or scoping a new initiative, one question cuts through the noise faster than any other: do we trust the data this would run on?
Not "do we have data." Most organisations have plenty of data. The question is whether that data is clean enough, governed enough, and understood well enough that you'd stake a business decision on it.
If the honest answer is "not quite", that's the real project. Not instead of AI. Before it, and alongside it.
The good news is that building a solid data foundation and pursuing AI capability aren't separate workstreams. Done well, they're the same programme of work. The foundation informs where AI creates the most value. The AI use cases clarify which parts of the foundation need to be strongest.
You don't have to choose between moving fast and getting it right. But you do have to start with what's real.
“Before working with Flock, pulling together a clear financial picture across our businesses was a significant manual undertaking, one that could take days and still left room for inconsistency. Now, that information is available in near real-time, standardised across all 30 of our investment companies, and accessible to the people who need it. We’ve only just scratched the surface of what this data can do for us.”