AI Made Everyone a Builder. It Didn't Make Everyone an Engineer.

I tell every client the same thing: use AI, and get your team using it this week.

I've been building software for businesses since 2004, and this is the biggest shift I've seen. It's also giving some people confidence their experience hasn't earned yet, while leaving plenty of others not sure where to start.

We're earlier than it feels

Scroll LinkedIn and you'd think every business has AI running through its core. The data says otherwise.

A recent study of nearly 6,000 executives across the US, UK, Germany and Australia found senior leaders average about an hour and a half of AI use a week. That's roughly 18 minutes a working day. Australia had the lowest business adoption of the four countries.

If leadership is spending 18 minutes a day with these tools, we're a long way from mature adoption across the rest of the business.

And the tools are the easy part. Most businesses carry years of legacy: spreadsheets that became systems, processes built around one person's knowledge, and platforms that were never designed to talk to each other. New tooling doesn't clear any of that away. The legacy has to move with it, or AI just speeds up the mess.

We've seen this movie twice before.

The first time: the cloud

Once, standing up infrastructure meant a sysadmin, a server room and a purchase order. Then AWS and Azure let anyone with a credit card do it in ten minutes. That was a real leap forward.

It also gave us open storage buckets, surprise five-figure bills, and servers nobody patched because nobody knew they existed. Standing it up was never the hard part. Keeping it secure, backed up and running at 2am was.

The second time: Agile

Agile promised faster delivery through short cycles. A lot of businesses heard "no more planning." They dropped the architecture, the governance and the documentation, and called the chaos iteration.

Agile never meant less discipline. It meant running the full lifecycle (plan, build, test, release) in smaller loops. Teams that understood that shipped faster. Teams that didn't shipped rework.

The third time: AI

Now anyone can describe what they need and watch it appear. That's a good thing, and we use these tools every day. But AI is a sharp knife. In your own kitchen, go for it. Once you're feeding customers, you need kitchen standards.

The day the laptop died

One of the most technically advanced businesses we work with found this out the hard way.

A key person had built something impressive: around 100 scheduled jobs running through an AI assistant, driven by spreadsheets, with business logic that only they fully understood and only they could access. It worked well, right up until their laptop died.

Operations stopped. Nothing could run until the laptop was restored and everything recovered.

You can guess what we were building straight afterwards: proper automation, on dedicated infrastructure, with the logic documented and owned by the business rather than one person's machine.

Nobody in that story did anything wrong. They built something valuable on top of legacy spreadsheets. It just never made the jump from prototype to production.

The Laptop Test

Before your business comes to depend on something you've built with AI, ask one question: if this laptop died tonight, would the business stop tomorrow?

If the answer is no, keep going. It's yours: your notes, your drafts, your personal workflows. Experiment, move fast, learn.

If the answer is yes, it's time to operationalise it. That means infrastructure it doesn't share with someone's inbox, security and access that don't depend on one person, monitoring that tells you when something fails, and logic written down where the business can see it.

Build it to learn. Build it to last.

This is the part that excites me. For as long as I've been in software, the hardest job has been getting the logic out of the expert's head and into something an engineer can build. Requirements got lost in translation, and projects delivered what the spec said, not what the business needed.

Now the person who knows the process best can build a working prototype themselves. That's the best spec we've ever been handed.

So build it. Prove the logic. Then bring in people who have spent decades making systems robust, scalable and maintainable to turn it into something the business can rely on. That's where our experience earns its keep: the questions nobody asks in the demo. What happens at ten times the volume? What happens when a connected system goes down? Who gets the alert? Who owns it when you're on leave?

Not at the prototype stage yet? That's most businesses.

Everything above assumes your team already knows how to use AI well. For most businesses, that's not the case yet, and that's fine. Remember the 18 minutes.

That first step matters just as much. Working out where AI fits in your business, getting your people confident with it, and deciding which legacy processes need to change before any tool will help. We do that work too. The businesses that get the foundation right are the ones whose prototypes are worth operationalising later.

Where your IT provider fits

Your MSP keeps your environment secure and running, and good ones are worth every dollar. Turning a prototype into a core business system is a different discipline: process design, integration architecture and software engineering. The best results come when both are at the table.

Wherever you're starting from

If you're not sure where AI fits yet, we can help you take the first step properly. If you or your team have already built something that's quietly become critical, bring us the prototype and we'll help turn it into something your business can run on.

Either way, let's talk.

Get in touch to book a conversation →

Source: Yotzov, Barrero, Bloom, Davis et al., "Firm Data on AI," NBER Working Paper 34836, 2026.