There is a reliable pattern in transformation work. A process is slow, so it gets digitised. The digital version is faster, but the same number of steps remain, several of which serve no current purpose. The organisation has made an inefficiency permanent and harder to change.
Simplification first is less satisfying than building something, and it consistently produces the better outcome.
The cost of automating the status quo
Every step in an automated workflow is something to configure, test, document and maintain. Steps that exist out of habit carry the same ongoing cost as steps that create value.
Worse, once encoded in a system, a redundant step acquires an air of authority. It becomes “how the system works” rather than a decision somebody made years ago for reasons that no longer apply.
Questions that surface unnecessary work
The second question is the most revealing. If nobody can describe a consequence, the step is a candidate for removal rather than automation.
- Who reads this output, and what decision do they make with it?
- What would happen if this step were skipped for a month?
- Is this approval managing a real risk, or a historical incident?
- Are we entering this information somewhere it already exists?
- Does this exist because of a system we no longer use?
Where the extra steps come from
Redundant work is rarely the result of carelessness. It usually accumulates through reasonable local decisions: a report added after a board asked one question, a second approval introduced after a single expensive mistake, a manual reconciliation created to bridge two systems that were later integrated.
Each made sense at the time. Together, and unreviewed, they form processes that consume significant capacity while nobody can explain their purpose.
Simplify, standardise, then digitise
Remove what is unnecessary first. Then agree a single way of doing what remains, because automating three regional variations of the same process triples the build and the maintenance for no benefit.
Only then introduce technology. By this point the process is smaller, clearer and cheaper to support, and the requirements are obvious enough that you are less likely to buy something oversized for the problem.
Expect resistance to removal, not to technology
In our experience the technology is rarely the contentious part. Removing a step is, because someone introduced it and it may still feel like a control they rely on.
The way through is evidence rather than argument: agree a trial period, measure what actually happens without the step, and decide from the result. Most trials end with the step retired and the person who introduced it satisfied that the risk was managed.
The compounding benefit
Simplification makes every subsequent decision easier. Clear processes are easier to document, easier to train new staff on, easier to measure, and considerably easier to automate well.
It is also the stage that makes AI genuinely useful later. A model or workflow applied to a clean, well-understood process delivers reliable value; applied to a tangled one, it produces confident output nobody trusts.