Free · by Timothy Gaull

The AI Foundation Kit

Get how your business actually runs out of your head and into something AI can use. You don't have to write it. The kit has AI interview you for it.

Your email is used only for the kit and, if you check the box, the follow-up and occasional notes. It's never sold or shared.

Free to use and share under CC BY 4.0.

How it works

How work moves through any business

Every business has a version of this. The kit maps yours by interviewing you about one real, finished job, then tags each stage by the kind of work it takes. Pick your business to see it in your terms.

  • Mechanical: AI can run it
  • Hybrid: AI drafts, you sign off
  • Judgment: Stays with a person
  1. Hybrid

    Request comes in

    Log it, check the basics

    e.g. A quote request by email
  2. Judgment

    Qualify

    Is this work we want?

    e.g. A site visit
    DeclinedLogged with the reason and date
  3. Hybrid

    Scope

    What exactly is the job?

    e.g. Measurements and photos
  4. Hybrid

    Price

    What will it cost and earn?

    e.g. An estimate
    LostThe quote didn’t win
  5. Mechanical

    Commit

    Signed, deposit in

    e.g. A signed quote and deposit
  6. Judgment

    Deliver

    Do the work well

    e.g. Build and install
  7. Mechanical

    Get paid

    Bill, collect, follow up

    e.g. The final invoice
Where work leaves is where the judgment lives. Examples are illustrative, not a real company’s process.

What changes

What happens to each stage

The tag on each stage decides its future. Routine stages become skills your AI runs. The calls only you can make stay with you.

  1. Hybrid

    Request comes in

    Log it, check the basics

    Skill + your OKAI drafts, you approve
  2. Judgment

    Qualify

    Is this work we want?

    YouYour call, now written down
  3. Hybrid

    Scope

    What exactly is the job?

    Skill + your OKAI drafts, you approve
  4. Hybrid

    Price

    What will it cost and earn?

    Skill + your OKAI drafts, you approve
  5. Mechanical

    Commit

    Signed, deposit in

    SkillRuns on its own
  6. Judgment

    Deliver

    Do the work well

    YouYour call, now written down
  7. Mechanical

    Get paid

    Bill, collect, follow up

    SkillRuns on its own

2 stages run as skills

Written once, tested on your past decisions, checked every week. You stop doing them by hand.

3 stages get a skill and your sign-off

The AI does the legwork. You make the one decision inside it and approve.

2 stages stay with you

The map records what you look at and why, so your calls get faster, and one day teachable.

Your week moves toward the calls only you can make.

Why it compounds

Good work feeds the next job

Each stage hands its output to the next, and the last one asks for the review that brings in the next request.

  1. Hybrid: Request comes in
  2. Judgment: Qualify
  3. Hybrid: Price
  4. Mechanical: Commit
  5. Judgment: Deliver
  6. Mechanical: Get paid
  7. Mechanical: Ask for a review

Reviews bring the next request

Each stage feeds the next

The output of one stage is the input of the next: a request becomes a qualified job, then a price, then a signed commitment. That is how skills chain together.

The routine stages keep turning

Commit, get paid and ask for a review run as skills, so the loop keeps moving while you are on the job.

Close the loop on purpose

The review request is one small skill. It feeds the top of the loop, where new customers find you.

Why it works

You can't automate what isn't written down.

Most AI adoption stalls for the same reason. The tools are fine, but every chat starts cold, because the real process, the pricing rules and the judgment calls live in the owner's head. This kit builds that missing foundation first, then turns it into AI skills one at a time.

Every week, steps 3 and 4 repeat

  1. Map

    Two or three 45-minute interviews about real, finished work.

    You getA stage-by-stage map
  2. Context

    What you carry in your head, written down once: what you take on, your pricing assumptions, what done means.

    You getA context pack
  3. Skills

    Your busiest routine stage becomes a skill, tested on five of your own past decisions before anyone trusts it.

    You getA tested skill
  4. Bank corrections

    Every week, each place a skill got it wrong becomes one dated edit.

    You getA log that proves it improves
  5. Teach

    New hires make the call first, then compare it with the skill’s reasoning.

    You getA team that learns your judgment
At 90 daysA map of how work actually moves. A context pack that ends the re-briefing. Two or three verified skills with dated changelogs. And the real test: someone other than the owner can run one workflow to the owner’s standard.

In practice

The bottleneck is rarely the people

Three real projects. Each fix moved routine work into a simple tool the team already uses, and gave their time back to the work that needs them.

A fabrication shop

Before

Calls missed while the owner is on a job site in the busy season.

After

A request form and a virtual assistant catch leads while the owner is on site. He calls back the ones worth quoting.

Owner’s time goes to site visits and accurate estimates.

An animal rescue

Before

Adoption applications on paper, and paper gets lost.

After

Volunteers list pets from their phones, and applications arrive complete, online.

Volunteer time goes to the animals.

A global sales team

Before

Years of client relationships kept in private spreadsheets, off the system.

After

A relationship stage with no pressure to close, so the whole picture is finally visible.

Advisors’ time goes to the relationships.

The routine moves to simple tools. The judgment stays human, and gets more time.

What's inside

Six plain files, plus a Claude Skill.

Everything is plain text, so it works in any AI tool today and in whatever you use in two years.

01 · Service delivery map

The interview. The AI asks you one question at a time about real jobs you have done, and writes down how work actually moves through your business, stage by stage.

02 · Context pack

The facts every stage assumes but nobody wrote down: what you take on and what you turn away, your standing assumptions, your shorthand, what “done” means.

03 · Skill extraction

Turns one mapped task into a written skill an AI can run and a new hire can learn from, in one fixed format so skills chain together.

04 · Example: quote request triage

A fictional worked example for a small fabrication shop, showing what a draft skill looks like, holes and all.

05 · Verification and retro

How to test a skill against five of your own past decisions before you trust it, and the 15-minute weekly habit that makes it better.

06 · Example: deal intake

A second fictional example, for a real estate investor screening new deals.

How to start

Which AI do you use?

Pick yours. Each way takes about two minutes, and the kit runs on your own account, so your business details stay with you.

  1. Turn on code execution

    In Claude, open Settings, then Capabilities, and switch on Code execution and file creation. Skills need it.

    Claude's Capabilities settings with the Cloud code execution and file creation switch turned on, labelled Required for skills.
  2. Upload the zip as it is

    Open Customize, then Skills. Click Add, then Upload skill, and choose the zip you downloaded. Don’t unzip it.

    Claude's Skills page with the Add menu open, showing Upload skill, Create a skill and Create with Claude.
  3. Check it’s there

    You’ll see ai-foundation-kit with its switch on. Click Try in chat, or start any new chat.

    The ai-foundation-kit skill page in Claude, marked New, with a Try in chat button and its switch turned on.
  4. Start the interview

    Say “Start the AI Foundation interview.” Claude loads the kit and asks one question at a time. Stop anytime; next time, say “Let’s keep going.”

    A Claude chat showing Loaded skill, then Claude explaining the AI Foundation Kit and its five steps.

Stuck, or want someone to run it with you? See how we’d work together →

A preview

The eight rules the AI interviewer follows.

These come straight from file 01. They're what make the difference between a useful map of your business and a generic chat.

  1. One question at a time. The AI waits for your answer. It never sends a list of questions.
  2. Real work, past tense. "Walk me through the job you finished last month" gets the truth. "What's your process?" gets the version nobody follows.
  3. Stop at every soft verb. When you say "I look at it and decide," that's hidden judgment. The AI asks what you looked at first and what would have changed your answer.
  4. Numbers, word for word. Prices, thresholds, lead times and limits are written down exactly as you say them.
  5. Ask for the exception. Every rule gets a follow-up: "When do you break that?" The exceptions are where the expertise lives.
  6. Never invent. Anything the AI inferred rather than heard is marked [UNCONFIRMED]. A visible hole beats a hidden guess.
  7. Record first, improve later. Fixes go in a separate list, so you describe the process instead of defending it.
  8. 45 minutes, then stop. Tired answers are vague, and vague answers make bad AI skills.

Questions

Common questions

What is the AI Foundation Kit?

A free set of plain-text files that turns any AI chat tool into an interviewer. It asks a business owner how work actually moves through the business, writes down the rules and numbers only the owner knows, and turns the most repetitive task into a tested AI skill.

Who is it for?

Owners and operators of businesses from one person up to roughly 200, such as trades, shops, manufacturers, professional services firms and real estate investors, where most decisions still run through the owner.

How do I teach AI how my business works?

Write down what only you know, then give it to the AI every time. The kit does this by interviewing you: it maps how work moves through the business, collects standing facts like what you take on and turn away, pricing assumptions and approval gates, and turns one recurring task into a skill that is checked against five past decisions before you trust it.

Which AI tools does it work with?

Claude, ChatGPT, Gemini, Copilot or any AI chat tool that accepts a long message. In Claude it also installs as a skill.

What does it cost?

Nothing. It is licensed CC BY 4.0, so you can use, adapt and share it, including with clients, as long as you credit Timothy Gaull.

How long does it take?

Two or three 45-minute interview sessions produce the first map. Most owners have a first tested skill within a few weeks.

Why do you ask for my email?

To keep the download from being scraped. If you check the box, you'll also get the follow-up on how owners are using their first AI skill, plus occasional notes. Your email is never sold or shared, and you can unsubscribe anytime.