The Steps That Exist for No Reason

Ask why a step exists and you will hear one of two answers. Somebody explains the reason, or somebody says that is how we have always done it.

The second answer is worth more than the first, and it is the cheapest improvement available to most businesses.

Where useless steps come from

Nobody adds a pointless step. Every one of them was rational when it was introduced, and the reason expired without the step being removed.

A mistake somebody made once. A costly error in 2021 produced a check. The check now runs on every job forever, at a cost far exceeding the original error, and the person it was aimed at left two years ago.

A client who is gone. One demanding client required a specific format, a weekly report, an extra approval. They left in 2019 and the process still does it for everyone.

A system you no longer use. Data entered in a particular order because the old software demanded it. New software does not care and the habit persisted.

A person who wanted visibility. A manager asked to be copied on something. They were promoted, moved, or left, and the step remains.

A rule somebody misremembered. Frequently a regulation that was never quite what people think it was, or that changed years ago and nobody re-read it.

The cost is measurable

The Canadian Federation of Independent Business, in the January 2025 edition of its Red Tape Report, found that businesses spent 768 million hours on regulatory compliance, equivalent to nearly 394,000 full-time jobs, and that the smallest firms pay over five times more per employee than the largest.

That report measures government regulation rather than self-imposed process, and it says nothing about automation. What it establishes is the scale of administrative load on smaller businesses, and it is worth pairing with an uncomfortable observation: a meaningful share of the burden inside most companies was not imposed by anybody. It was added internally and never reviewed.

Your own red tape is the part you can remove this month without asking permission.

Automating one is the expensive mistake

This is why the article exists. The worst possible outcome of an automation project is a fast, reliable, well documented version of a step that should not happen.

Now it costs money to run, it is harder to remove because somebody built it, and it has acquired legitimacy. A step performed by a person can be dropped in a conversation. A step encoded in a system requires a project.

Always ask whether to remove it before asking how to automate it. In most process reviews we run, somewhere between a fifth and a third of the steps do not survive that question.

How to find them

Map the process as it actually happens, then go through it step by step with one question: what would break if we stopped doing this.

Three answers and three responses. A specific consequence somebody can name means keep it. Nobody knows, or the reason turns out to have expired, means stop it. And it is a legal or contractual requirement means check the actual document, because this is the answer most likely to be wrong.

That last one is worth its own note. Requirements get remembered as stricter than they are, and rules get carried forward for years after they changed. Somebody should read the contract or the regulation rather than repeating what everybody believes it says.

Stop it, do not debate it

The fastest method, and the one that produces an answer rather than a discussion.

Stop the step for a month and see if anything happens. Tell people it is a trial and that it will be reinstated if something breaks.

This works because arguing about a step in a meeting is unwinnable: nobody can prove it is unnecessary and everybody is cautious. A month of evidence settles it. Most steps stopped this way are never reinstated and nobody ever mentions them again.

Where AI changes the analysis

Two things, and the first is the one businesses never manage on their own.

AI shows you the process as it is, not as documented. Reading across your systems, timestamps and correspondence produces the actual path work takes, including the steps people skip, the ones they do twice, and the informal ones that appear nowhere. Every process map drawn in a meeting is a description of the official version, and the gap between that and reality is where the waste lives.

AI measures what each step costs. How long work waits at that point, how often the check catches anything, how many times the approval was ever refused. That last figure ends most arguments on its own: an approval step that has not rejected anything in two years is not a control, it is a delay.

Then the steps that survive are worth handing to AI, and they will be a shorter list than you started with.

The steps you should not remove

A caution, because this exercise builds momentum and momentum removes things it should not.

A control that has never caught anything may be working precisely because it exists. People behave differently when they know a check happens, and removing it changes the behaviour it was quietly shaping. That is genuinely different from an approval nobody has ever refused.

The test is who knows about it. If the people doing the work know the check happens, it may be deterring something and the absence of findings is not evidence of uselessness. If it happens invisibly, downstream, and has caught nothing in two years, it is a delay.

Anything touching safety, money leaving the business, or a legal obligation gets the benefit of the doubt and a proper look rather than a month-long trial.

Do this before any automation project

• Take one process and list every step as it actually runs.

• Ask what breaks for each one, and write the answer down rather than accepting a nod.

• Check the requirements that are claimed to be legal or contractual against the document.

• Stop three steps for a month and see what happens.

• Then automate what is left.

An afternoon, and it routinely removes more time from a process than the automation that follows it.

Sources

• Canadian Federation of Independent Business, Canada’s Red Tape Report, 7th edition, January 2025.

AI Optimize maps how your process actually runs before proposing to automate any of it, and will tell you when the right answer is to delete a step rather than build one. That work sits under Workflow Automation.