Day 007 The Foundation Run

Workflow 007: The Templates Skill

Part of Scale You Workflows: 31 real workflows from my business, so it runs without you. Prompts included. Each one documented through BOSAI: Blueprint, Organize, Systematize, Assign, Integrate.

A skill that holds your proven templates plus examples of past wins, so AI can produce a customized proposal or email that already sounds like you.

Time to build
1 hour
Difficulty
Beginner
Tools
Claude, Zoom
Le-an Lai Lacaba Creator of BOSAI. Ten years placing executive assistants with founders.

What if you never wrote the same document twice?

Not a blank template you fill in. A skill that already knows how you write proposals, what you charge for, and what the good ones looked like.

A templates skill holds your recurring documents (proposals, invoices, the emails you send constantly) along with real examples of past versions that worked and the variables that make each one different. You give it the context for a new client, and it produces a customized draft that sounds like you. Mine runs on Claude and takes its input straight from the discovery call transcript.

The part that makes it work isn't the template. It's the examples.

What this replaces

Rebuilding the same document for every client, with small custom changes each time.

I run three businesses, and the AI systems agency is the one where this hurt most. Every proposal needed to be genuinely custom. One client is a startup, the next has been running for ten years without ever building their systems. Different sizes, different problems, different tools.

So I'd open the last proposal, strip out the previous client, and rebuild it. Same shape every time, an hour of work every time, and the quality depended on which old proposal I happened to open.

Blueprint

Give the skill the context for a new client. Get back a customized document.

Trigger: manual, when a document is needed. Inputs: the template, examples of past versions that worked, and the variables for this specific client. Output: a drafted document, customized, ready for a human to review and send.

For proposals, the variables that actually change the document are:

  • How big the company is
  • What problem they're actually solving for
  • Which tools they already use
  • Their timeline, and what's driving it

That last one matters more than people expect. A founder growing the business needs a different proposal than a founder documenting everything because they're preparing to sell. Same service, completely different framing.

Organize

What has to exist first:

  • The template itself, in its current best form
  • Three to five real past versions that worked, with the client details stripped out
  • The list of variables that make each version different
  • A source for the input, usually the call transcript
  • Your voice rules, or a pointer to your Me skill

The past examples are the whole difference. A blank template produces generic output. Examples of what actually worked, plus the reasoning behind the choices, produce something that sounds like you and lands like the ones that closed.

Systematize

Save this as a skill, one per document type.

You are my [DOCUMENT TYPE] writer. When I say "draft a [DOCUMENT TYPE]
for [CLIENT]," you produce a customized draft.

What you have:
- My current template, attached
- Examples of past [DOCUMENT TYPE]s that worked, attached
- My voice and writing rules

Before writing, gather these variables. Pull them from the call transcript
if I've given you one. Ask me for anything missing rather than assuming:
- [VARIABLE 1, e.g. company size]
- [VARIABLE 2, e.g. the problem they're solving for]
- [VARIABLE 3, e.g. tools they already use]
- [VARIABLE 4, e.g. timeline and what's driving it]

Then write the draft:
- Follow the template structure exactly
- Customize the framing to their situation, not just their name
- Match the tone of the past examples, not a generic business voice
- Where past examples handled a similar situation, follow that approach
- Never invent prices, timelines, deliverables, or commitments. If it
  isn't in the template or the transcript, ask.

After the draft, tell me:
1. Which past example you leaned on most and why
2. Anything you had to assume
3. Anything missing that I should fill in before sending

Save it as a draft. Never send.

How to set it up

  1. Pick the document you write most often.
  2. Gather three to five past versions that worked. Strip out client details.
  3. Write down the variables that made each version different. This is the step people skip and it's what makes the output custom instead of generic.
  4. Save the template, the examples, and the prompt together as one skill.
  5. Run it on a real client. Correct the output, then put the correction into the skill file.

How do I get the input without an intake form?

From the call. The information was already said out loud on the discovery conversation, so point the skill at the transcript instead of asking a client to fill in a form.

Assign

What the AI does: gather the variables, choose the right approach based on past examples, write the draft, and flag anything it had to assume.

What stays human: the input at the start, and the judgment at the end. Someone has to read the draft and decide whether it's something you'd be proud to send. That's not a formality. It's the entire quality control.

There's a discipline attached to that review, and it's where most people lose the compounding.

When something in the draft is wrong, fix the file, not just the draft.

It's tempting to correct the output, send it, and move on. Do that and you'll correct the identical thing next month. Feed the correction back into the skill and the problem is gone permanently. That's the difference between a tool that stays flat and one that gets better every client.

Rule of thumb from this build: document what already works, and tell the AI why it worked.

The why is the part everyone leaves out. Handing over five old proposals teaches format. Explaining which ones closed and what made them land teaches judgment.

Integrate

The templates skill doesn't sit alone:

  • The call transcript feeds it, so nothing gets retyped
  • It reads the Business skill for offers and pricing, and the Me skill for voice
  • Once a document goes out, a follow-up loop watches for silence. Mine flags anything unanswered after a few days and drafts the nudge
  • Every correction goes back into the file, so the next one starts better

Mine also does something I didn't plan for at the start. Because it has seen which clients worked out well and which didn't, it can tell me when a new one looks like a poor fit. Not a decision I hand over, but a useful second opinion from something that has read every proposal I've ever sent.

Tomorrow: the brain dump sorter. What to do with the pile of thoughts you emptied into a note at midnight.