The second automation: what to pick once the first one runs
The first project teaches you the method. The second is where companies either build momentum or quietly stop.
Say you followed the advice from February: one process, finished, running for a month. Supplier invoices, most likely. It works, the exceptions are known, the person who checks it has stopped dreading Mondays. What now?
This is the decision point where most SME automation programmes end, not with a failure but with nobody choosing the next one.
The wrong second project
Something ambitious, to prove the concept scales. It does not need proving. You proved it. An ambitious second project has a high chance of failing and taking the credibility of the first one with it.
Something in a different department, to spread the benefit. Tempting and premature. You have one person who now understands how this works. Sending them into an unfamiliar process doubles the learning cost.
The right second project
Pick the process immediately adjacent to the first one. If you automated the reading of supplier invoices, the natural next step is matching them against orders, or the approval routing, or the payment file.
Three reasons this works.
The data is already clean. You deduplicated the suppliers for project one. That work is done and it carries over entirely.
The same person owns it. They know the exceptions, the vocabulary and where the bodies are buried.
The value compounds. Two connected steps automated end-to-end save more than twice what one saves, because the handover between them disappears too.
The test before you start
Ask the person who runs the first automation what annoys them about it now. Not what is broken — what is merely irritating.
The answer is almost always your second project, stated precisely, by the person who will benefit from it. In one SME it was « I still have to open the order in the other system to check the price ». That sentence is a complete specification.
And when to stop
After three or four adjacent steps, the remaining work in that area is judgement rather than repetition. Stop there and start again somewhere else, with one process, finished, run for a month.