Automating order tracking: 12 hours of work

Starting point: a small online retailer tracks orders in a spreadsheet updated by hand three times a day, by someone who could be doing something else. Goal: a dashboard that updates itself, with no manual routine left. Here is how the 12 hours break down, structured the way an applied-AI track in the catalogue would frame it.

Hours 1 to 2 — Map the current process. Before automating anything, list precisely every existing manual step: where an order arrives, who enters it, how often, and above all which errors keep recurring (duplicates, missed entries, wrong category). Automating a poorly understood process only reproduces the errors faster.

Hours 3 to 5 — Connect the data source. Set up automatic order collection (form, sales-platform API, or a regular file import) into a structured store. This is the most technical step, but also the most mechanical: once the connection is in place, it needs no further intervention.

Hours 6 to 8 — Processing rules and alerts. Add the business logic: automatic duplicate detection, order categorisation, and the metrics that matter (average processing time, overdue orders). This is where AI comes in — to sort and flag, not to decide in place of the person responsible.

Hours 9 to 11 — The dashboard itself. Build the view that replaces the spreadsheet: today's snapshot, the week's trend, visual alerts on overdue orders. The success criterion is not aesthetics — it is that a non-technical person understands the dashboard in ten seconds.

Hour 12 — Real-world testing and documentation. An automation that breaks silently after a month is worse than a reliable manual routine. The last hour goes into checking that the system flags its own failures, and documenting in one page how to restart it without depending on a single person.

Result: a task that took about 20 minutes, three times a day, six days a week — nearly 6 hours a week — becomes a two-minute check. The 12-hour investment pays for itself in a little over two weeks.