I'll start with a confession. We caught it.
Not long after the good AI tools arrived, 2Gen came down with a solid dose of AI fever. For a moment there, we genuinely thought we could take over the world. Every process looked like a candidate. Every meeting ended with someone saying "we could get AI to do that". We were going to automate everything, probably by Friday.
We got better. Most businesses do. But it's worth knowing what it looks like, because if you haven't had it yet, you probably will.
The symptoms
Here's a quick self-check. One or two of these is normal. Five or more and you should lie down somewhere quiet.
- Every meeting ends with "could AI do that?" Including the meeting about where to go for lunch.
- Someone has built a chatbot. Nobody uses it. The person who built it doesn't use it either.
- You pay for more AI subscriptions than you can name.
- A process nobody has looked at in five years is suddenly "ripe for automation".
- The word "agent" no longer makes you think of real estate.
- You've started a sentence with "Imagine if…" and finished it with "…the whole business".
- The plan has a logo but not a problem.
What it looked like here
At the peak, the list of things we were going to build was longer than the list of things we actually needed. Some of what we built was pretty sweet. Some of it wasn't that flash. A few were genuinely crappy.
My favourite flop wasn't even for work.
I asked Claude to put together a chores chart for our rural lifestyle block. It came back looking great. Everyone had a proper job. Everyone except my youngest daughter, who had been given one task: pick up sticks.
Nobody asked for that. Claude decided she needed something to do and filled the gap all by itself. To be fair, on a lifestyle block there are always sticks.
It's funny on a chores chart. It's a lot less funny anywhere else. AI is very good at sounding sure of itself. When it doesn't have the information, it will often make up something plausible rather than ask. Do that in a quote, a policy or a client's network settings and nobody's laughing.
That's why everything we build now works from real information, like your own documents and systems. It's also why someone always checks the output before it matters.
None of that was wasted. It's how you learn. But it's an expensive way to learn if you don't stop and take your temperature.
The cure
There's no pill. What worked for us was slowing down and following four rules.
1. Start with the problem, not the tool
"We should use AI" isn't a problem. "Our team spends every Monday morning re-keying the same orders into two systems" is a problem. If you can't say who it helps and what it saves, put the shiny thing down.
2. Pick the right processes
Not everything is worth automating. A job that takes forty minutes eleven times a year will cost you more to automate than it ever saves. We wrote a whole article on the work not worth automating, because we learned that one the hard way.
3. Do it safely
When you're excited, security feels like the boring bit you'll sort out later. Later never comes. Work out what the AI can see, who it can act for and what gets logged before anything goes live. Fixing that after everything is running is miserable.
4. Take controlled steps you can measure
Build one useful thing. Check your team actually uses it. Check it saves what you hoped. Then build the next one. It's slower in week one and much faster by month six, because you're not unpicking the projects that went nowhere.
Is it contagious?
Very. It spreads fastest through conference talks, LinkedIn and sales demos.
You'll know it's under control when:
- you can name the problem before you name the tool
- you know what each AI tool can see in your business
- you can point to what something saved, in hours or dollars
- your staff use the thing without being reminded
Where we are now
We still get the odd flare-up. Someone sees a new tool and the old twitch comes back. The difference is we now have a process for it. The process wins.
Today we have our own software factory running, building tools that make our business better. Every one of them went through those four rules first.
If your business has a dose of AI fever, that's fine. It means you're paying attention. Before you build anything, start here. We'll help you work out what's worth doing and what isn't.