Which process should you automate first?

Once you start looking, you see work that could be done better everywhere. The challenge is no longer finding candidates, but choosing. And most businesses choose wrong, because they go for the process with the biggest theoretical payoff.

That is exactly the process that sits deepest in your business, has the most exceptions and touches the most people. Your first project needs to prove something else: that this works, here, with our people.

Score your candidates on four axes

List your candidates and give each a score from low to high on these four points. You are not looking for the highest total score. You are looking for a specific profile.

• Frequency: how often does it happen. High is good.
• Predictability: does it usually go the same way. High is good.
• Number of systems involved: how many packages need to talk to each other. Low is good for a first project.
• Damage when it goes wrong: what breaks if it fails. Low is good for a first project.

The profile you are after: frequent, predictable, few systems, low damage. That is usually not your biggest problem. That is exactly the point.

What that looks like in practice

For an installation company it is often the shared info mailbox: having incoming messages sorted and labelled so a fault report doesn’t get lost among the newsletters. Few systems, high frequency, and a mistake is a wrong label that someone corrects.

For a wholesaler it is often reading in supplier documents: packing slips and price lists that are now retyped by hand, turned into a proposal that someone approves. For a haulier it is usually status information to customers, so the “where is my order” phone call disappears. For a bookkeeping firm it is preparing and recognising the documents clients send in.

What these cases have in common: they are visible. Within two weeks someone on the shop floor notices that things are different. That is worth more than a saving that only shows up in the quarterly figures.

Order of play in the first weeks

1. Follow the process once in full, from the moment it comes in until it is done, and count the actions and the waiting time. Don’t ask how it goes. Watch.
2. Measure one number you want to see again later. Number of emails that ended up in the wrong place, days between work done and invoice sent, hours of retyping per week.
3. Build the narrowest piece that has value on its own. Not the whole process. One link in the chain.
4. Let it run alongside the old way for two weeks. You can only compare if you see both.
5. Measure again and decide. Expand, adjust or stop. Stopping is a valid answer.

An example of how narrow you should start

Take the shared mailbox of an installation company. The big wish is: process everything automatically into work orders in the schedule. The narrow first step is much smaller. Incoming mail gets a label: fault, quote request, invoice, supplier, other. Nothing else happens. People work the same way they always have, but see at a glance what is urgent.

That can be built within two weeks, it touches one system, and a mistake is a wrong label that someone corrects. If it works, you take the next step: forwarding faults automatically to the planner. If it doesn’t work, you have lost little and you know exactly why. That order feels too cautious, and it is the reason such projects actually get finished.

What goes wrong in this phase

The exceptions are always more numerous than everyone thought. You hear that a process “basically always” goes the same way, and in the first two weeks six cases come along that didn’t fit the story. That is normal and it is no reason to stop, but it does mean your schedule needs slack.

The second thing that goes wrong is the handover. A process that works fine as long as the person who built it keeps an eye on it falls over as soon as that person is gone. Make sure someone in your own business understands what happens and where the controls are. That takes a few hours of explanation you would rather not schedule, and that you earn back later.

And sometimes the conclusion is that the process itself is no good. Then automating is the wrong answer and redesigning how the work is done is the right one. That is cheaper, but it is a harder conversation.

When you need us and when you don’t

If you can already see clearly which process it should be, and everything sits in one package that simply has the feature built in, go ahead without us. Many first steps are a matter of switching on settings and an afternoon of trying things out.

We help when you can’t choose between ten candidates, when the process runs across several systems that don’t talk to each other on their own, or when you want an outside pair of eyes that dares to say your candidate is a bad idea. A half-day core session is usually enough for that.