Article
If AI feels like magic inside your business, you may be starting on the wrong side of the problem.
Cookies by George asked me to fix its bulk-ordering process. Customers download a template and upload roughly a hundred recipient orders at a time. Processing one of those orders takes the team two to three hours. When Christmas gets busy, those hours are hard to find.
I had said we would use AI, but I did not begin with AI. I began by mapping what happened.
Where did customers go? What did they download? What did they upload? Where did the process break?
Putting those steps on paper made the work visible. Once you can see a process, you can touch it, question it, and measure it. That gives you something much more useful than a vague request to “add AI.”
Fix the clear problems first
The map showed several ways to make the process more efficient without AI. Automatic address checking could catch predictable errors. A proper web service could carry the data. A better interface could make the job easier for customers, while better admin tools could help the people processing the orders.
Those are ordinary software and workflow improvements. A clear process lets the system do the same dependable thing each time without asking a model to make judgment calls.
AI came at the end, where the inputs stopped being predictable. Sometimes a customer deletes a column. Sometimes they enter strange data that does not belong. A deterministic system cannot always repair those exceptions, so that is where AI can help.
Put one workflow on paper
Pick one workflow in your business. Order intake and onboarding are both useful places to look. Write down where people go, what they receive, what they return, and where the work stalls or breaks.
Keep it simple. The first pass does not need to be clever or polished. It needs to make the real work visible.
You may find that address checks, a better interface, or more useful admin tools solve much of the problem. Then you can put AI to work on the odd cases that remain, with a clear job and a process around it.