Every founder I talk to says some version of the same sentence: my work is too custom to automate. I used to nod along, because for a long time I believed it. Now I ask them to send me their last five proposals. Four of those five usually share the same skeleton, and the founder has never noticed, because they rebuild it from scratch every single time.
That is not a custom business. That is an unexamined one.
Custom usually means unexamined
I have been teaching since 1998 and running businesses since 1995, and I have heard some version of "my work is too custom" from photographers, agency owners, accountants, retreat operators, interior designers, and a wedding planner who does four events a year at $200k each. The industries share nothing. The sentence is identical, word for word.
What they mean is that the finished thing feels different every time. And it does. A brand audit for a law firm does not read like a brand audit for a skincare company. Fine. But the process that produced both of them is nearly the same process: the same discovery arc, the same six categories in the scope doc, the same access requests in the kickoff email, the same three questions that surface the real problem.
The deliverable varies. The decisions behind it barely do. When founders say the work is too custom to systemize, they are describing the surface and skipping the process.
Draw the repeatability map in one sitting
The map is three columns and ten examples. Nothing fancier. If you want to know how to systemize a service business, this is the first hour of the work, and everything else gets easier after it.
Open a Notion table or an Airtable base. Three columns: Repeats, Variable, Judgment.
Then do this:
- Pull your last ten deliverables. Not your best ten, your last ten. Recency beats pride.
- Give yourself 90 minutes in one sitting. This is a sorting exercise, not a research project.
- For each section of each deliverable, mark it. Repeats means the words are basically the same. Variable means different client, different language, same structure. Judgment means a human had to decide something based on context.
- Count. Most service founders land between 55 and 70 percent repeat. In four years of doing this with people, I have never seen it come in under 40.
That percentage is your automation budget. Not your whole business. The number tells you how much of your week is running the same play, and it usually surprises people, because the pain never felt repetitive. It felt like drowning.
What's becoming clearer to me is a simple rule: automate the repeat, protect the judgment. Because of that, the map is not really an efficiency exercise. It is a decision about what you are for.
What to hand to AI first, and in what order
Order matters here. Hand off by risk, not by annoyance. The thing that annoys you most is often the thing with the most judgment inside it.
Move one: capture the intake. This is the highest-value, lowest-risk handoff in almost every service business. Record the discovery call with Fireflies or Otter, then run the transcript through a saved prompt in Claude or ChatGPT that outputs a structured brief with fixed headings: what they said, what they meant, what has to be true for this to work, open questions. You read it and correct it. You never send the raw output.
Move two: the first draft of the scope doc. Set up a Claude Project or a custom GPT and load it with your three best past proposals. Ask for a draft against the brief. The draft comes back at maybe 60 percent, and 60 percent of a document you used to write from a blank page is a real afternoon back. Fireflies into Zapier or Make can wire the transcript to the doc automatically so the file is waiting for you before you sit down.
Move three: the recap and status email. This is where founders quietly lose four to six hours a week. Build a snippet library in TextExpander or Notion, one snippet per recurring situation, then let AI fill in the client specifics. It reads like you because you wrote the bones.
That is what to automate first in a service business. Not the fee negotiation. Not the crisis call. The capture, the draft, and the follow-up.
What to keep bespoke forever
Three things stay human, and I will fight about all three.
The pricing conversation. You are reading whether they are stretching, whether a partner needs convincing, whether the number lands as relief or as a wound. AI cannot hear that, and if you let it try, you lose the signal that tells you whether this relationship has legs.
The moment something goes wrong. A missed deadline, a bad edit, a launch that flopped. The apology and the recovery plan are the relationship. Automate that and you have told the client what they are worth to you.
The read on the room. The scope creep that is really a budgeting problem. The client who says everything is fine and is not fine. The project that has quietly changed shape. That is the part people pay you for, and it does not live in a template.
Bespoke is not the words on the page. Bespoke is the judgment underneath them. The people who scale do not automate judgment, they protect it, and they stop spending it on status emails.
If the idea of separating judgment from repetition is new for you, this comparison is a useful place to see how founders structure the work once the map is drawn.
Standardize custom deliverables without making them generic
The fear is real: automate the proposal and every client gets the same document, and the thing that made you good disappears. That happens when you build a template. It does not happen when you build blocks.
A template has blanks. A block has a locked paragraph and two or three variable slots. The difference matters, because the locked paragraph carries your thinking and the slots carry the client's specifics.
How to build it:
- Take the ten deliverables from your map and find the twelve sections that repeat most often. Twelve is usually enough to cover 80 percent of what you write.
- Write each block once, in your voice, with the strongest version of that argument you have ever made. This is the highest-value writing you will do all year.
- Leave slots for exactly four things: the client's name, the problem in their own language, the number, and the deadline. Nothing else gets a slot.
- Store the blocks in a Notion database or an Airtable base with a status field, and pull them into Google Docs or your proposal tool from there.
- Test each block by swapping the client name only. If it still makes sense, it was a block. If it breaks, it was judgment, and it goes back in the human column.
A photographer does this with a shot list. An accountant does it with a month-end close pack. An agency does it with an audit. Same move every time: standardize custom client deliverables at the block level and let the assembly stay bespoke.
When founders tell me they are drowning in admin and need to clone themselves, this map is usually the first thing I hand them. Not a software stack. A sorting exercise. If you want to do it with people around you instead of alone on a Tuesday night, the workshop is where we build these live.
Because of that, the goal was never to make your work less custom. The goal is to stop spending your best hours retyping the parts that repeat, so the custom part gets the version of you that is actually rested and paying attention. That's the whole point.
If this resonates and you want to build this kind of infrastructure in your own business, the Mastermind is where we do the work live. You can learn more at mastermindshq.business.