The work your team does because the software will not

At 8:40 every morning someone on your team opens a dashboard. They scan four or five numbers, decide that nothing looks wrong, and close it. Twice a month something does look wrong, and they message a colleague about it. That person has been doing this for two years. It is not in their job description. It has never appeared in a budget. If they left tomorrow, the first anyone would notice is when a problem sat for three weeks with nobody watching.

Every company has a layer of this work. It is invisible for a simple reason. It produces nothing you can point at. It generates no document, no revenue, no ticket. It exists only to move a fact from one place to another place, because the software that already held the fact was never going to do anything with it.

The five shapes it takes

You can recognise this work by its shapes. There are only a few.

Checking. A person opens something on a schedule to see whether anything happened. Dashboards, inboxes, folders, reports that get exported on Mondays.

Watching a line. A person keeps a number in their head and acts when it crosses a threshold. Stock below a level, a queue above a length, a balance under a floor.

Chasing a status. A person asks another person or another system where something is. Did that ship, did that get approved, did that customer ever reply.

Copying. A person reads a value in one system and types it into a second one, because the two do not talk and never will.

Relaying a failure. A person notices that something broke and tells whoever can fix it. The noticing is the entire job. The fixing belongs to someone else.

None of these are hard. That is exactly why they hide. Nobody escalates work that takes four minutes. They just do it, every day, for years.

The exercise

Take one week. Ask everyone on the team, including yourself, to write down every recurring thing they do that begins with noticing something.

The test for whether an item belongs on the list is the first verb. If the task starts with open, check, scan, read, watch, glance, or ask, it goes on. If it starts with build, write, decide, call, or design, it does not. You are not cataloguing the work. You are cataloguing the sensing.

For each item, record five things. What triggers it. What the person looks at. What they do next. How often it happens. What would go wrong if nobody did it for a month.

Then add the two entries that matter most, and take your time with them. First, the moment the fact became true somewhere in your systems. The payment failed at 21:14. The stock level crossed the line on Thursday afternoon. The customer’s card expired on the first of the month. Second, the moment a person acted on it.

The difference between those two is the gap. Write it down for every item on the list.

Sort by the gap

Now sort the list by that gap, longest first. This is the part that changes how you see the whole thing.

The instinct is to sort by minutes spent. That is the wrong measure, and it is why this work stays invisible. The morning dashboard check costs four minutes. Four minutes is nothing. But the fact it was checking may have been true for sixteen hours, and on the days something was wrong, that sixteen hours is the entire cost of the item. The person’s time is trivial. The delay is not.

A sorted list usually reveals three bands.

At the top sit items where the gap is days and the action takes seconds. A supplier confirmed a delay on Tuesday and three teams found out on Friday. Someone hit a limit over the weekend and heard from you on Monday. These are the most expensive lines on the page and they are almost always the least discussed, because the work involved is so small that nobody thinks of it as work.

In the middle sit items with a short gap and high frequency. Someone is effectively on call for a system that could tell them directly. The cost here is not delay, it is that a person’s attention is pinned to a screen instead of to the work you hired them for.

At the bottom sit items where the gap barely matters and nothing depends on the answer. Some of these should not be automated. They should be stopped. If nobody can say what would go wrong in a month without them, try a month without them.

Judgement, and things that look like judgement

Two piles. Real judgement stays with a person. Relay work does not.

Four questions sort them.

In the last fifty times this happened, how often did the answer differ from the obvious one? If the answer is always the same, there is no judgement in it. There is a rule that nobody has written down.

Could the person write the rule they follow in one short paragraph, and would a colleague reading that paragraph reach the same answer? If yes, it is a rule.

Does the person need something that lives outside every system you own? A phone call, a relationship, the knowledge that this particular customer had a bad month. That is judgement, and it should stay human.

Is the action reversible? Retrying a card is reversible. Cancelling a contract is not. Where undoing is hard, keep a person in it even when the rule is clear.

Here is the part that most lists get wrong. Almost every item splits in half. There is a noticing half and a deciding half, and they are not the same work. A refund above a certain amount is a real judgement call and should stay with a person. But nobody needs to be the one who spots it. The system knew the moment it arrived. A person sitting in front of a screen so that they can be told something a machine already knew is doing relay work, even when the decision that follows is entirely theirs.

So separate them. Keep the deciding. Give up the watching.

We built our own Shopify apps, xupsell and xbundle, on that principle. No merchant is expected to check a screen to find out that something happened in their store.

What the list is worth

Nothing on this list is new work. It has been running for years, spread thin across everyone, which is precisely why it never got examined. Written down in one place and sorted by the gap, it becomes something you can talk about. You can point at the top three lines and say what each one costs. You can hand it to a new person and stop pretending the knowledge lives anywhere but in one head.

Most of it will turn out to be relay. A smaller part will be real judgement, and that part is worth protecting.

If you make the list and want a second pair of eyes on it, book a call.