Before you automate it, ask whether it should exist at all
Automating a bad process gives you a faster bad process. In thirty years of changing how organisations work, the biggest savings have nearly always come from the process, not the technology.
The trap
Someone spends a day a week copying figures from one system to another. The obvious answer is to automate the copying. The better question is why the figures need to be in two places.
Sometimes the answer is good: the accounts system and the operations system genuinely need the same information. Often it is history. Somebody once needed a report, the report became a habit, and the habit became a job. Automate it and you have spent money to preserve a habit.
Where the cost really hides
Two pieces of work from long before AI show the pattern.
A government records operation had to absorb a rise from around one million files to seven or eight million. The instinct was to treat it as a storage problem: more space, more shelves. When we followed the work, storage was not where the money went. The cost sat in movement: retrieving files, moving staff between sites, the distances travelled for every request. The answer was to consolidate the warehouses and price the service on what delivery really cost. Nobody would have found that from the invoice.
A bank ran mail and document processing across scattered locations. Each site worked, in its own way. Redesigning the process around two centres, built so that either could cover for the other, meant the operation ran with around 95 people instead of more than 200, and became more resilient at the same time. The technology helped, but the saving came from redesigning how the work flowed.
The same thing happens in a ten-person business. The cost is rarely in the step everyone complains about. It is in the handoffs between steps: waiting, re-checking, chasing, and the same information typed twice.
Four questions for every step
- Remove: what happens if we stop doing this? If nobody notices for a month, stop.
- Simplify: can it be done with fewer steps, fewer people or fewer systems? Every handoff is a place for delay and error.
- Combine: are two people doing parts of the same job without knowing it? It is more common than anyone admits.
- Then automate: only what is left, and only where it saves more than it costs.
A useful test. Write the process down as it is actually done, not as the manual says. Count the steps and mark every one where information is re-typed, re-checked or waited for. In most processes we review, those marked steps are a third or more of the total, and almost none of them add anything the customer would pay for.
What it looks like in practice
A business shipping frozen food decided every day, by judgement, which boxes needed early delivery or extra ice. The question was not how to automate that judgement. It was whether the decision needed judgement at all. It did not. The forecast, the box size and the delivery window were all known in advance, so the rules were written down and applied to every order, with a sentence explaining each decision.
Sometimes the answer is no software
We say this in every review: if an existing product already solves the problem properly, buy it. If the process should simply stop, stop it. A review that ends with "do not build anything" is still worth paying for if it saves you building the wrong thing.
Have a job like this? Tell us in a few lines what gets copied, checked or chased by hand. We will come back with questions, not slides.
Book a callNext: What is an MCP connector, and why would a business want one?