Automate where the work waits

Have you put an AI tool into the business this year, set it up properly, and watched nothing actually get faster?

The tool works. The emails draft themselves, the meeting notes write up, the quotes come out in half the time. And the customer still waits just as long for an answer as they did before you bought it.

That is not a bad tool. It is a tool pointed at the wrong place.

Most owners automate the task they find slowest, or the one they find most annoying. Neither is usually where the time goes.

Most businesses I look at are paying for at least one tool like this.

Automate the wait, not the work

Here is what most people miss. A piece of work spends very little of its life being worked on. Most of its life, it is waiting — in an inbox, in a tray, on someone's list for tomorrow.

An enquiry arrives. It sits. Someone forwards it. It sits again.

It goes to whoever does the pricing, and waits there too. The real work across that whole fortnight might add up to under an hour.

And every business has one step that everything queues in front of. That is the constraint.

Speed up a step before it and you only feed the pile faster. Speed up a step after it and that step stands idle, waiting for the pile to clear.

You did not make the business faster. You made the queue longer.

So choose the target before you choose the tool. Finding it takes an afternoon:

1. Pick one real piece of work. An enquiry, an order or an invoice from this week — not a made-up example.
2. Follow it end to end. Write down every handover: who had it, when it arrived with them, when they started, when it left.
3. Split the clock in two. Time being worked on in one column. Time waiting in the other.
4. Circle the longest wait. That handover is your first automation. Not the task everyone complains about.

What goes there is usually unglamorous.

A rule that routes the enquiry without anyone forwarding it. A draft prepared before the person opens the file. An alert when something has sat too long.

The clever use case saves minutes. The boring one saves days.

There is one honest exception. Sometimes the longest wait sits in front of a person's judgement — a price only you set, an approval only you give.

That is the one thing you should not automate. Hand judgement to a machine and you get wrong answers faster.

Delegate it instead: write down how you decide, give the call to someone else with a limit, and then automate around them.

Point the tool at the slowest task and the pile grows. Point it at the longest wait and the pile moves.

Toyota built machines that stop

Taiichi Ohno joined Toyota in 1943 as a shop-floor supervisor and rose to vice president. The production system he helped build, and described in his 1978 book, treats waiting as waste in its own right. It sits on his list of seven wastes alongside overproduction, excess inventory and defects.

The second pillar of that system is older than the car company. More than a century ago, Sakichi Toyoda built a power loom that stopped itself the moment a thread broke, instead of weaving on and turning out defective cloth. Toyota calls the idea jidoka — "automation with a human touch."

On a Toyota line today, a machine that detects something abnormal stops. An operator who spots a problem can pull a cord and stop the line. A board lights up to show exactly where, and people go to that spot to fix the cause.

Because the machines watch themselves, one person no longer has to stand and watch one machine.

Toyota did not win by automating more than everyone else. It won by automating in a way that exposed problems instead of hiding them.

Most businesses automate to keep things moving, and a pile of finished work in front of a stuck step looks exactly like productivity. Toyota's machines were designed to stop and point at the stuck step.

Where I learned to look

I have worked this way for years.

It came from the boardroom more than the shop floor. I chaired the board of an ASX-listed company, and a board teaches you where time actually goes.

Very little of it is spent deciding. Most of it is spent waiting for a decision to be ready — for a paper, a number, a signature, the next meeting.

Nobody on a board is slow. The work just sits between people. Once you have watched that from the chair, you stop asking how long a task takes and start asking where it sits.

That is the first question I bring to any process now, well before anyone mentions software. The answer is almost never the step the team complains about. It is the handover nobody owns.

Do this before Friday

Take one enquiry that arrived on Monday. Follow it through the business until it is answered, or until Friday, whichever comes first, and note every time it changed hands.

On Friday, add up two numbers: hours it was worked on, hours it waited. Circle the single longest wait.

That is where your first AI tool goes. Not the job you hate. The place the work sits.