20 Trees, Two Tonnes of Rubble and a Lesson About AI

Discover the vital groundwork needed for successful AI implementation, much like planting trees for lasting growth.

Article

I've just finished planting 20 trees. Not the little ones you pick up from a garden centre, either. These were 13-foot monsters that arrived on the back of a lorry.

The goal was simple enough: create a hedge that would give my family privacy for years to come, while doing something positive for the environment. The reality, however, turned out to be considerably more complicated than I'd imagined.

Before planting anything, I had to think about the long-term impact of the roots on nearby foundations, roads and surrounding infrastructure. Then came the physical work of digging three-foot holes in stubborn clay, in blistering heat, while pulling out bricks and rubble the builders had conveniently buried years earlier.

My neurodivergent brain immediately labelled it *legacy tech debt*. Yes, that's genuinely how I think.

I ended up shifting around 60 bags of rubble, roughly two tonnes, over 20-odd trips to the tip in my car. After a considerable amount of lifting, protein shakes and Monster drinks, the ground was finally ready for the next stage.

That meant sorting the drainage, mixing in compost, stinky cow manure, bark and horticultural stone, and making sure the roots had the right conditions to establish themselves. I wasn't interested in cutting corners. After all, I don't have a slush fund sitting around to replace 20 failed trees.

And the work didn't stop once everything was planted. Each tree needed around two minutes of watering every day, which quickly becomes another job when you're looking after 20 of them. Eventually, I set up an AI-powered irrigation system that I could control through my iPhone, giving me more time with my five-year-old and one-year-old.

Somewhere between hauling rubble, improving the drainage and setting up the irrigation, it struck me how similar the whole experience was to introducing AI into a business.

## Everyone sees the finished hedge.

Very few people see what it took to get there.

It's much the same with GPT, Claude and other AI tools. People see an impressive demonstration, watch a system complete a task in seconds and understandably wonder why implementation can't be just as straightforward. From the outside, it can look almost plug-and-play.

What they don't see is the groundwork. The risk assessments, the old problems that need clearing out, the effort involved in understanding existing processes, mapping workflows and connecting new technology to systems that may have been in place for years.

Just as I couldn't simply dig a shallow hole, drop in a tree and hope for the best, a business can't always introduce AI into an existing workflow without understanding the conditions surrounding it. Poor-quality information, unclear responsibilities, disconnected applications and processes that haven't been reviewed in years can all undermine what initially looked like a promising solution.

And then there's everything that happens after implementation. Employees need time to adapt, results need measuring, problems need fixing and the technology has to become part of everyday operations rather than another system nobody quite knows how to use.

None of that happens without the right people, either. I'm fortunate to have some excellent specialists in my network who understand that the less glamorous work is often what makes the real difference. Their experience matters because a technically impressive solution isn't necessarily one that can be implemented, maintained and relied upon.

I think this is where businesses can underestimate the investment required for AI. Getting something running is one thing. Making it reliable, useful and commercially worthwhile over time is quite another.