Everyone wants to automate everything. Start with the 1 thing that is bleeding.

01The rule that decides the order

Pick the leak that is obvious, that you can count, and that gives a result the same week. Not the biggest saving on paper, not the most interesting build. The one you can point at on Friday and say that took 4 hours before and 11 minutes today.

The first automation has to earn the right to build the second one.

The reason is about people, not tech. You have to convince a team, a client, or the part of you that thinks this is a distraction from real work. A win you can prove buys everything after it. A 6 week project that saves 20 hours a month on paper buys nothing until it lands, and most never land.

02Automate client intake first

Here is the leak, and almost nobody sees it happen. A new lead calls or fills in your form. It lands in somebody's inbox. The person who can answer is with a client, in court, on site. By the time anyone replies, the lead has hired whoever answered first.

Speed to lead decides more deals than quality does. Unfair, and still true. People hire the firm that replies, not the best firm, and that goes double when the problem is urgent or emotional. If people buy under pressure, this is where your money goes.

That last line is where people get it wrong. The temptation is to let the system reject anyone who fails your rules. Do not. Your rules are a good guess, not a law of nature. Wrongly rejecting 1 good client costs far more than a 30 second look.

The real value is not the software. It is writing down your judgment about what makes a good client. Nobody buys that off a shelf. In the builds I do now it sits in the command center, the AiOS, so the answer is the same whether the owner is awake or asleep.

03Then documents and proposals

The second one is the most wasteful thing in most service businesses, and everyone treats it as normal. Expensive people putting contracts, engagement letters, proposals, reports and filings together by hand from templates. Seniors whose time is worth far more, or juniors who should be learning something else.

Intake data flows into the right template, and the sections get picked by the type of work. A draft appears in minutes instead of hours, ready for a human to read and approve. You are not replacing judgment, you are removing typing. A 40 minute engagement letter becomes a 3 minute read, and across every job that is an extra person of capacity with no salary.

Your team stops typing and starts advising. Nothing leaves without a signature.

That is what makes careful, old school businesses say yes. Full control, and only the grunt work goes. In the Dubai work I do, a Telegram assistant earns its keep here. The team sends a voice note from a phone, and the document gets built and filed in the right place.

04Then follow up and collections

2 different jobs, same reason they come 3rd. Both are things humans avoid because they feel awkward, not because they are hard. So the automation gets welcomed instead of resented. Nobody defends their right to chase invoices personally.

Follow up is where most of the money in a service business already sits, in people who were interested and never got chased properly. Collections is the other half. Firms are brilliant at earning money and terrible at collecting it. The invoice goes out and sits, because asking for payment feels beneath everyone, and the debt ages while the business looks profitable on paper.

A system that tracks every invoice, sends polite and relentless reminders in your voice, and escalates on a clear path takes the awkwardness away. Nobody has to be the bad guy. Getting paid faster across the whole book is a real cash swing, and cash is oxygen.

05Then reporting, and only then

Reporting comes 4th and it hurts, because it is the 1 thing founders want first. A dashboard feels great, demos beautifully, and on its own changes nothing.

Build it once the work underneath is real. A dashboard on a broken process only tells you faster that the process is broken. When you do build it, keep it to the handful of numbers you will actually act on.

It is the same mistake I see all over the market. Everyone bought the tool and nothing changed in how the work gets done. I wrote about that in most firms bought Ai and few changed how the work gets done. The dashboard first is the same disease.

06Put a number on the bleeding before you pick

The order above is the general answer. Your order comes from simple maths, and it takes 10 minutes.

For each job write down the hours a week and multiply by 4.3. Say intake admin eats 6 hours a week. That is about 26 hours a month, most of a working week, gone, on details somebody could have typed in themselves.

Now price it. Not at what you pay an assistant. At what an hour of the person doing it is worth. If that is a founder hour, be honest, not modest.

2 things fall out of a table like that. The biggest number is rarely the job that annoys you most, and annoyance is what most people automate first. And the total is usually frightening, which is why you write it down.

Do the 1 at the top. Not all 4, and never reporting because it is fun to build.

07What to leave alone

This half of the question gets no attention and it is where the real damage happens. 3 things I would not automate, and 1 I would gate.

People think I mean a machine could not attempt it. The judgment is the product. Automate the assembly around it and leave the decision alone.

Automating a relationship moment is easy to do and expensive to your name. People forgive slow. They do not forgive feeling processed.

A process nobody has written down becomes a system that locks in the confusion and runs it faster. Write it on paper first. You will usually find the process was the problem.

The gate is not a lack of trust. It is what lets everything else run free. Tools only matter after these lines are drawn, and even then only together. I listed mine in the 5 Ai tools I pay for.

08Do not build all 4 at once

You will be tempted, because the full picture is exciting and each piece makes the others better. Resist it for the same reason you would not hire 4 people in 1 week.

You cannot tell what is working or what broke. When something goes wrong you are debugging 4 systems and the 4 joins between them at once.

Build 1. Prove it with real cases. Then the next. The order is slower on paper and much faster in real life.

If you take 1 thing: automate where the money is visibly leaking and where you can prove the fix inside a week. In nearly every service business that is intake, and it funds everything after it, in belief as much as in cash. Then keep going in order, with a person on every decision your clients pay for.

09Questions, answered