Before you automate it, try to delete it.

When a new capability lands, everyone asks what they can automate. The sharper first question is what they can stop doing. Automating work that should not exist just makes the waste cheap, permanent and invisible.

A finance team I worked with last year wanted to automate a report. Weekly, twelve pages, out the door every Monday morning. They had costed an automation that would build it overnight and have it sitting in the right inboxes before anyone arrived. Sensible enough, on the face of it.

So I asked who actually reads it.

Long pause. The honest answer was that one person skimmed the first page, and the rest was archived unread. We had been about to spend good money making something nobody wanted arrive faster.

This happens more often than you would think. When a new capability lands, the instinct is to point it at whatever is in front of you. There is a queue of tedious, repetitive work in every organisation, and AI is very good at tedious and repetitive. So the first question everyone asks is "what can we automate?"

I think that is the wrong first question. The first question is "what can we stop doing?"

Automation makes everything run faster, including the waste

Most processes in a mid-sized business were never really designed. They accreted. Someone needed a report in 2019, the person who needed it moved on, and the report kept going out because nobody ever said stop. Multiply that across a decade and a couple of restructures, and you end up with a lot of motion that is no longer attached to a decision.

Automating that motion does not fix it. It industrialises it. You take work that quietly wasted a little time each week and you make it run overnight, every night, with nobody watching. The cost of producing it drops, so the scrutiny drops with it. The waste becomes cheap, permanent and invisible, which is arguably worse than the manual version. At least the manual version annoyed someone enough that they might eventually question it.

Subtraction is unglamorous. There is no demo for deleting a report, no vendor pitch for the workflow you switched off. But it is the cheapest improvement available to you, and it is the one move the technology will never suggest on its own.

What we actually do first

Before we automate anything on an engagement, we walk the process and ask a few blunt questions:

• Who reads this, and what do they do differently because of it? If the honest answer is "nothing", we stop here.

• Does this step exist for a reason that is still true, or for a reason that left the building three years ago?

• If we switched it off for a month, who would actually notice? (You can run that experiment. The results are usually clarifying.)

• Are we about to automate the workaround for a problem we could simply fix at the source?

What survives that is a much shorter list. It is also a better list, because everything on it has earned its place. Now automation is worth the effort, because you are pointing it at work that genuinely needs doing, not at a habit the organisation has quietly outgrown.

None of this is an argument for going slow on AI. If anything it is the opposite. The teams getting the most out of these tools are not the ones who automated the most tasks. They are the ones who were honest about which tasks were worth keeping. Delete first, then automate what is left.