Author: Rod Fowler
From mainframes to machine learning
I’m not an AI guru, and I’m suspicious of anyone who claims they are in a field that changes faster than the average software update.
What I do have is 50 years of hands-on experience adopting new technology – from mainframes in an IBM software lab in Stockholm to the smartphone that now tells me where to drive, when to stand up, and which meetings I’m already late for.
I started my career before mobile phones, the internet and personal computers were part of everyday life. I could drive a car one‑handed while reading a printed map (unsafe, but impressive) and I had about 50 landline numbers memorised.
Today, I let GPS do the navigating, and my phone remembers everyone’s number while I struggle to remember my own PIN.
In all that time, I have never seen a technology as powerful – or as quietly terrifying – as this current wave of AI.
AI is the new electricity
When electricity first arrived, it didn’t turn up with a manual, a standards body and a friendly YouTube explainer. There were very few uses for it at first, almost no safety guidelines, and plenty of painful “learning experiences”.
Over time, we built things around it:
- Appliances that made it useful.
- Massive infrastructure to generate and distribute it.
- Safety standards, fuses and rules so people could use it without getting fried.
AI today feels a lot like those early days of electrification.
It’s touching every industry and every part of our personal lives.
There’s an explosion of new uses, applications and “must‑try” tools, and the infrastructure behind it – data centres, software, chips – is being rolled out at high speed.
The big difference?
Electricity was easy to see, hear and feel.
If you wired something badly, sparks flew, smoke escaped, and people screamed.
With AI, a lot of the damage can be silent, invisible and hard for non‑experts to notice until it’s too late.
The washing machine that went rogue
When the electric washing machine turned up, the user interface was pretty simple:
You threw in dirty clothes, added powder, pressed a button, and 45 minutes later you could see and feel the result. Clean clothes. Happy days. Very low likelihood of your washing machine secretly emailing your boss.
With AI, it can be more like this:
You load your clothes, press “start”…
and the machine quietly connects to every other appliance in your house, your neighbour’s water supply and a random laundromat across town to “optimise” the cycle.
Then it tips a mixed pile of clothes onto your bed and asks you to decide, on the spot:
- Which items are actually yours.
- Which ones are clean.
- Which ones shrank.
- And which ones belong to a stranger but are now mysteriously part of your load.
The interface looks simple and friendly – one text box, one button, maybe a smiling robot icon. Under the hood, there are complex connections, hidden risks and outcomes that are much harder for a non‑expert to judge.
And unlike your washing machine, this thing might have access to customer data, financials, internal documents and even the ability to take actions on your behalf.
That should get your attention.
Your staff are already using AI (whether you like it or not)
Here’s the part that should make business leaders sit up straight:
Unlike PCs or early mobile phones, which cost thousands of dollars, anyone can access surprisingly powerful AI tools for free or for the price of a couple of coffees a month.
That means:
- Employees are experimenting with AI tools on their own initiative.
- Many are pasting in emails, reports, proposals and even customer details to “save a bit of time”.
- A lot of this is happening on personal accounts, outside your line of sight and certainly outside your IT policies.
In other words, your organisation may already be deeply plugged into AI – you just don’t know which socket, which voltage or whether the wiring complies with anything remotely resembling a safety standard.
When the powerful tools have the weakest protection
A recent AI Risk Quadrant report looked at 100 popular AI agents and scored them on two things:
- How exposed they are: their “Attack Surface” (what they can access, what they can do).
- How well protected they are: their “Defence Controls” (security, guardrails, checks and balances).
A disturbing number landed in the worst possible place: high attack surface and low defence.
To translate the technical jargon into plain language:
- The tools with the most power often have the least protection.
- The tools with the strongest protection are often the least capable or least attractive to users.
- Only a small minority – around one in ten – are both powerful and well‑defended.
So if your people are simply grabbing whatever looks shiny from the AI tool shelf, the odds are not in your favour. There’s a good chance they’re choosing something that is more “bare wires in a bucket of water” than “properly installed safety switch”.
This should be frightening news to any employer with no AI policy or controls in place.
Why this matters for leaders (not just IT)
This is not just a “cybersecurity problem” or an IT issue.
It’s a leadership issue.
AI tools are increasingly able to:
- Read and summarise sensitive documents.
- Draft emails, proposals and contracts that go straight to customers.
- Trigger workflows, actions and decisions inside other systems.
If you don’t set clear boundaries, your people will make it up as they go along.
Not because they’re reckless, but because they’re busy, the tools are helpful, and the marketing pages don’t exactly lead with “By the way, here’s how we might accidentally leak your data.”
Ignoring AI is not a strategy.
Your staff are already using it.
Your choice is whether they do that in a controlled, safe environment – or in the AI equivalent of a backyard powerboard taped together with duct tape.
What responsible AI adoption looks like
The answer is not to ban AI and pretend it’s a fad that will go away.
Electricity didn’t go away either, and nobody suggests we go back to candles and hand‑washing our clothes (I’ve seen enough laundry to know that’s a non‑starter).
