Where does the mess come from?
Most organisations that start with automation hear the following at some point during the project: 'We need to get our processes in order first.' Sometimes that is correct. But it is also used as an excuse to do nothing. To determine who is right, you need to answer one question: is the mess structural or operational? Structural mess means that people handle every case differently, that no consistent process exists, or that exceptions are in the majority. Operational mess means that the process does exist, but that execution is slow or error-prone due to manual work, rekeying data, or the absence of proper tools. Only structural mess makes 'cleaning up first' worthwhile. With operational mess, automation is the solution.
When do you genuinely need to clean up first?
There are situations where automating without preparation backfires. The clearest one: when nobody agrees on what the correct process is. If three planners each follow a different approach and you automate one of them, you have not resolved the disagreement, you have frozen it. The same applies when source data is unreliable, not because it is entered poorly, but because there is no agreement on what should be entered at all. A system running on that data produces outcomes nobody trusts. In those cases, the honest conclusion is: the preconditions for automation are not yet in place. That is not a failure, it is a diagnosis. Address those preconditions first, and then automation becomes meaningful.
When does 'cleaning up first' become a mechanism for delay?
More often than expected, 'cleaning up first' is used to justify postponement. Not deliberately, but the effect is the same. The pattern looks like this: the organisation knows the process is not ideal, someone suggested that processes need to be in order before automating, and so an improvement programme starts that runs for months without anything changing in day-to-day operations. What is actually happening is that the process is recognisable and consistent enough to automate, but uncertainty about the investment or the change leads to delay. If an order processor does the same thing every day, reading a PDF, copying data into a system, and sending a confirmation, that process is in order. It is simply manual. That is exactly what you can address right away.
What do you gain by automating immediately?
When a process is consistent but manual, automating it immediately delivers three things. First, the error-proneness of rekeying and manual handoffs disappears. Second, you gain visibility into exceptions, because an automated system makes explicit what it cannot handle. Those are precisely the cases that deserve attention. Third, you move faster without adding headcount. In the current labour market, that is not a luxury but a necessity. Practice also shows that automation clarifies the process. Because you have to capture it in rules, implicit agreements surface. Many organisations discover what their process actually is during the automation project, not before it.
How do you assess this as an operations manager?
A simple test: take five random cases from the past month and walk through them with the people who handled them. If the steps are recognisable and the deviations are explainable, the process is ready for automation. If the steps differ fundamentally from case to case and deviations are the norm, there is work to do first. That work does not have to be a large improvement programme. Sometimes it is an afternoon conversation, a clear decision tree on paper, and alignment within the team. After that, automation can begin. At Bonsai we see this regularly in the operations of clients in logistics, trade, and manufacturing. The tendency to wait too long costs more than the imperfections in the process. An AI Worker that handles eighty percent of cases autonomously and routes the rest to a person is always better than a manual process that delays one hundred percent of cases.