Most companies pick the wrong process

Starting today, I am sending you the full method I use to make companies AI-native. Seven steps, one email at a time. This is step 1.

Before we start, one promise. This series is not a list of tools. It is a method. Seven steps, each one something you can do this week, in the order that actually works. By step 7 you will have one process running on AI, and you will know how to do the next one without me.

Let's go.

The most expensive mistake happens on day one

Almost every company starts its AI journey the same way. Someone senior gets excited, a meeting is called, and the team picks the most impressive process to start with. The complex one. The one that touches five departments. The one that would look great in a board update.

Six weeks later there is a demo that almost works, a team that quietly stopped believing in it, and a conclusion that "AI is not ready for us yet."

AI was ready. The process was wrong.

I have watched this happen more times than I can count, and I did it myself when I started automating my own business. My first instinct was to build something clever. What actually worked was embarrassingly boring: automating the replies to inbound sales inquiries. Same questions, same structure, every single day. That one process went from about 30 minutes per reply to 3. Not because the AI was smart, but because the process was simple, repetitive, and nobody enjoyed doing it.

That is the pattern. The right first process is boring.

The three-point test

Score every candidate process from 1 to 5 on each of these:

1. It repeats.
Daily or weekly, same shape every time. Frequency is not a nice-to-have, it is how the system learns. A quarterly process gives you four chances a year to see what breaks. A daily one gives you twenty a month. You will fix in three weeks what would otherwise take a year.

2. It has clear inputs and outputs.
An email comes in, a quote goes out. A document arrives, data gets entered. A call ends, the CRM gets updated. If you cannot describe the process in one sentence with a clear "in" and a clear "out", it is not first. It might be third or fourth, once you have the reps.

3. Nobody loves doing it.
Chasing overdue invoices. Copying data from one system into another. Drafting the same follow-up for the tenth time this week. Handing those to AI creates zero resistance in your team, because you are taking away the part of their job they complain about. Handing over the work people actually enjoy creates plenty of resistance, and resistance kills AI projects faster than any technical problem.

A process that scores 12 or higher out of 15 is a great first pick. Anything under 9, park it.

Ten candidates that usually score high

To make this concrete, here are the processes I see winning as a first pick, across almost every type of company:

→ Replying to inbound inquiries and quote requests
→ Drafting quotes from a standard template
→ Following up on unpaid invoices
→ Updating the CRM after calls and meetings
→ Preparing a briefing before a customer meeting
→ Chasing missing documents (signed contracts, proof of delivery, paperwork)
→ Answering the same 20 customer questions
→ Turning meeting notes into action lists
→ Compiling the weekly numbers report
→ Sorting and routing the inbox

Notice what is not on the list: strategy, pricing decisions, anything creative, anything a customer sees without a human check. Those come later, or never. That is fine.

The three traps

Trap one is the impressive trap I described above. Trap two is the "everything at once" trap: six half-finished experiments running in parallel, none of them owned by anyone. Trap three is the "tool first" trap: someone buys a platform, then goes looking for a problem. All three end the same way. One process, end to end, then the next. That is the whole method.

Your exercise for this week

Grab a note and write down the three most repetitive processes in your company. Score each 1 to 5 on the three points. Add them up. The highest score is your first process.

Here is the worksheet, copy it as is:

Process: ____________________
Repeats (1-5):            __
Clear in/out (1-5):       __
Nobody loves it (1-5):    __
Total:                    __ / 15

Do this for three processes. It takes ten minutes. That list is the start of everything that follows.

Honest cost

This step costs you ten minutes and a bit of ego, because your first AI process will not be the one you brag about at dinner. It will be the one that quietly gives your team a few hours back every week. That is the point. The bragging comes at process five.

In step 2 I will show you how to map your winner so precisely that an AI system can actually take it over. That step is where most do-it-yourself attempts die, so do not skip it.

Which process won on your list? Hit reply and tell me. I read every answer, and if you are not sure between two, I will tell you which one I would pick.

Talk soon,
Robin

PS: Rather talk it through than figure it out alone?