Everyone uses AI — few make it actually do the work

Almost every business owner we talk to already "uses AI." Opens ChatGPT, asks something, copies the answer. It's useful. But it's not a strategy — it's a faster Google search.

The difference between companies where AI actually saves money and time, and companies where it just looks modern, isn't the model or the subscription. It comes down to one thing: whether AI is pointed at an actual process, or just sitting open in a separate tab.

The two different things we call "using AI"

1. AI as a tool you ask. You open a chat, type a prompt, read the answer, copy it somewhere. Great for writing copy, summarizing, a headline idea. It saves minutes. You do the connecting by hand, every single time.

2. AI wired into a process. A form on your site → AI classifies the inquiry → drafts a reply → notifies the right person → logs everything to a sheet. Nobody opens a chat window. It just happens, every time, the same way.

The first is a convenience. The second actually changes how many hours you work a week, or how many inquiries turn into sales. Most companies stop at the first, because the second requires something chatbots don't hand you on their own — direction.

Why most rollouts don't go further

• There's no clear process to automate. If the work is already chaotic, AI just automates the chaos faster — it doesn't fix it.
• The data is scattered. AI can't "know" your stock if it lives in a spreadsheet on someone's desktop.
• Nobody tracks the result. A tool launches, gets used enthusiastically for two weeks, then quietly disappears because it isn't connected to anything else.
• The expectation is "magic," not "a system." AI doesn't replace a missing process — it speeds up one that already exists, at least on paper.

That's why so many AI pilots die quietly: a technical problem got solved that wasn't the real problem.

What it looks like when it works

In the ERP we built for Studio Botema, the automation isn't "an AI chatbot bolted on top" — it's built into the order flow itself: status, stock, supplier, all in one place, no chat window, nobody asking anything. In PropIntel, the AI property valuation isn't a conversation — it's part of the product, triggered automatically by a request and returning a result in seconds instead of hours of manual work.

In both cases, the AI is invisible to the end user. Nobody "chats" with it. It just does one specific, repeated task, wired into a real workflow.

Signs you're directing it, not just chatting with it

	An open AI tab	AI wired into the process
Who starts the action	A person, every time	An event (order, inquiry, appointment)
The data	Copied by hand	Connected automatically
What happens if you're away	It stops	It keeps working
Time saved	Minutes per request	Hours per week
Effect on sales/costs	Hard to measure	Measurable — less manual work, faster replies

If most of the left column sounds familiar, that's not a failure — it just means you're at stage one. The question is whether you want to stay there.

Where to start if you want it to actually work

1. Write down one process, not ten. The one you do most often and most mechanically.
2. Check where its data actually comes from. If it lives in three different systems, that's the first job — not the AI.
3. Decide what should HAPPEN automatically, not what you want to ask. That distinction is the whole thing.
4. Measure before and after. Hours, number of inquiries, response speed — whatever number mattered before the AI did too.

This is exactly the work I do as a consultant — I don't sell a tool, I look at the process before I suggest technology. Sometimes the answer really is automation. Sometimes it's just a better-organized spreadsheet.

---

Using AI but not seeing it move the needle on hours or sales? Tell us the process — we'll tell you honestly whether the problem is technical, or it's the process underneath.