In summary:
- Prepare the ground before planting the seeds: rushing to AI without sorting data foundations first doesn't accelerate progress, it accelerates poor decisions, because AI performs only as well as the data it is given.
- Good foundations are not a technical problem, they are an organisational one: the real work is agreeing definitions across teams, assigning clear data ownership, and building governance that gives people confidence to use data rather than reasons to work around it.
- The groundwork determines the outcome: a client who completed data foundations work before AI deployment identified over 200 hours of efficiency savings and moved from discovery to delivery in under a year, evidence that what happens before AI is switched on matters more than the tools themselves.
Last summer I decided I was going to grow tomatoes. I imagined delicious summer salads filled with ripe tomatoes and the quiet satisfaction of having grown something myself.
So, I bought tomato seeds. I filled some pots with compost. I watered them once and waited for something to happen.
Nothing did.
I'd assumed that seeds, soil, and a bit of water would be enough. It wasn't. What I hadn’t factored in was everything else tomato growing needs: regular attention, the right temperature conditions, and noticing early when something isn’t working.
What I ended up with was a collection of very dry pots and a firm lesson in the gap between starting something and sustaining it.
I've thought about my failed tomatoes a lot in the context of my work. Because the same pattern shows up again and again with AI, and most organisations are making exactly the mistake I made in the garden.
Organisations are planting the seeds, but very few are tending the ground.
The bit that gets skipped
Right now, a lot of leaders are under pressure to move with AI. To have a strategy, to show progress, and to not be the organisation still deliberating while everyone else appears to be growing something impressive. That pressure is real, and I understand it.
But the organisations I work with who rush straight to AI, without sorting their data first, don't move faster. They just get to the wrong answer more quickly.
They've planted the seeds, but they haven't prepared the ground.
Give it trustworthy data and it will help you make better decisions, faster. Give it data your teams don't agree on or can't access, and it will do exactly the same thing, just badly.
Here's a quick test. In your last leadership meeting, did two different reports give different answers to the same question? If so, that's a data foundations problem. And buying a new AI tool won't fix it.
It’s the equivalent of wondering why nothing is growing, without ever checking what’s happening in the soil.
I have that conversation more than any other. Not because organisations aren't investing in data, most are, but because investment in tools doesn't automatically produce investment in the basics underneath them. The dashboard gets built. The ownership and the definitions don't.
What good foundations actually mean
Data foundations isn't a technical term I use to make things complicated. It means something straightforward, can your organisation find, trust, and use its data when it matters?
In practice it means knowing who owns each data set, having definitions everyone agrees on so different teams asking the same question get the same answer, and making sure data is actually accessible without needing a specialist to translate it. Governance should make people more confident using data, not give them reasons to work around it.
What that looks like day to day is less dramatic than it sounds. A conversation about what a metric actually means before it goes into a report. A named person who is responsible when a data set looks wrong. A process for resolving disagreements about numbers that doesn't end in a spreadsheet stalemate. Small things, but they compound quickly, in both directions.
This is the tending. The unglamorous, ongoing work that doesn't get announced but determines whether anything you build on top of it actually grows.
What changes when you get this right
With a recent client, we helped the team go from a discovery engagement to a strategy implementation and delivery programme in under a year. In the process, we reviewed seven processes and identified over 200 hours of efficiency savings.
That didn't happen because we had better AI tools. It happened because the groundwork was done properly first. When the foundations are right, everything built on top of it works. When it isn't, nothing does, no matter how good the technology.
The other thing that changes, and this is harder to put a number on, is confidence. When leaders trust their data, they make decisions differently. They stop spending the first twenty minutes of every meeting arguing about whose figures are right, and start actually using them. That shift matters as much as the efficiency savings, even if it's harder to measure.
Where to start
My tomato plants didn’t fail because the seeds were wrong. They failed because I treated planting as the end of the work, not the beginning.
AI initiatives fail for the same reason.
If you want AI to deliver value, the most important work often happens before anything intelligent is switched on. Preparing the ground. Agreeing what good looks like. Checking early when something isn’t right.
That’s what allows anything else to grow.
On 13 May, I'm running a 30-minute session as part of our AI for Leaders series where I'll cover the most common data challenges I see slowing organisations down, what good foundations look like in practice, and five concrete steps you can take now, without waiting for a large transformation programme. If that sounds useful, I'd love to see you there.
Register here