The Default Shift: an AI mindset for your work

Time: about 10 min reading + 20 min practice

The gist

Imagine you're driving and suddenly notice that everyone is following new rules: the traffic lights are now run by AI and change three times faster. You can keep driving the way you always have. But whoever adjusts to the new logic gets there faster. The Default Shift is the moment you change your default (the setting you fall back on automatically) from "I'll do it myself" to "how can AI take on part of this?"

Key concepts

• Default Shift: a change in thinking where AI is your first resource, not your last
• The three M's: Mindset, Method, Machine: a three-level model of how your work changes with AI
• The 30% rule: before any task, ask "How can AI do 30% of this?"
• Dark Code: the danger of using AI-written code blindly
• Contractor Mindset: treating Claude as a contractor, not a fortune teller

Theory

The Default Shift: the main switch

🎨 Picture this: the Default Shift is like going from paper maps to GPS. At first it feels strange to let the app pick the route, but a week later you can't remember how you ever got around with a folded map in the glove box. AI becomes something that runs in the background, not a special occasion.

Most people use AI as a last resort: "I'll try it myself → it didn't work → I'll ask ChatGPT."

The Default Shift flips that around:

❌ Old default: "How will I do this?"
✅ New default: "How can AI do 30% of this while I think about strategy?"

This doesn't mean "hand everything to AI and forget about it." It means:

• Routine work (emails and messages, first drafts, research) → AI takes the first pass
• Strategy (what to build, why, for whom) → stays with you
• Control (review, corrections, the final call) → you're always the last line of defense

🎨 Picture this: a restaurant kitchen. The head chef doesn't peel the potatoes; the prep cooks handle that. The chef thinks through the concept of the dish, tastes, adjusts. But the chef plates the final element personally. AI is your prep crew.

The three M's: Mindset, Method, Machine

Adapting your work to AI happens on three levels:

M1: Mindset

🎨 Picture this: one cab driver thinks the job is turning the steering wheel. Another thinks the job is getting a passenger where they need to go, on time and without stress. To the first one, GPS looks like a threat; to the second, it's a helper. How deeply you understand the job decides how good your solution is.

The key question: what is your job, really?

If you think your job is "writing code" or "doing email marketing," AI looks like a competitor. If you understand that your job is "creating value for customers," AI becomes a multiplier.

Shifts in thinking:

• From "I do tasks" to "I run a system that does tasks"
• From "I need more time" to "I need to teach the system"
• From "this is my skill" to "this is my area of expertise, amplified by AI"

M2: Method

How your work processes change with the Default Shift:

M3: Machine

🎨 Picture this: at the first level, you bake bread yourself. At the second, you make the dough from a recipe. At the third, you run a commercial bakery: all you decide is which bread gets baked tomorrow. An AI OS (your own AI operating system) is your personal bakery.

The final level: you don't just "use AI," you have a working machine:

• Workflows (sequences of work steps) that run automatically
• Skills (reusable instructions) that hold your expertise
• Agents (programs that carry out tasks on their own) that take part of the work off your plate
• A knowledge base that grows with every project

Not everyone goes this far, and that's fine: the first two levels are enough for everyday work. The first-build module and the course library show how to put a machine like this together.

The 30% rule in practice

Before any task, ask yourself: "How can AI take on at least 30% of this?"

Examples:

Important: over time AI's share may grow, but start with 30% so you don't lose control.

Dark Code: the hidden danger

Dark Code is AI-written code that you deploy (put live, publish) without understanding how it works. If you don't write code, the same rule applies: don't put a spreadsheet formula, a calculation or a contract clause from AI into use if you can't explain it yourself.

The problem with this kind of code:

• It works only until the first unusual situation
• You can't debug it (find and fix the errors) when it breaks
• You don't know its limits
• You can't explain it to a client

🎨 Picture this: you took a pill from a neighbor "for a headache." The headache went away. But you don't know what was in that pill, whether you're allergic to it, or whether it's OK to have it with a glass of wine. This time you got lucky. One day you won't.

How to avoid Dark Code:

1. Ask for explanations: after each block of code, ask: "Explain what this function does in plain English"
2. Write tests together: have AI write a test for each workflow right away
3. Understand the architecture: even if you don't follow the details of the code, understand what the system does
4. Check edge cases: "What happens if the API (Application Programming Interface: the way one program requests data from another) doesn't respond? What if the data is empty?"

In the lessons where we build something, we go step by step, precisely so you understand every part.

Contractor Mindset: how to give Claude tasks the right way

🎨 Picture this: a contractor on a job site doesn't guess what the client wants. They ask specific questions: How many rooms? What materials? What's the budget? The more precise the spec, the more precise the house. Talk to Claude the way you'd talk to a contractor, not to a mind reader.

Claude isn't a fortune teller. It doesn't guess what you want. It's a qualified contractor.

Here's the difference:

❌ How people talk to a fortune teller:
"Do something about marketing for me"

✅ How people talk to a contractor:
"I sell AI courses for small business owners.
Audience: ages 35-50, small businesses, United States.
Write 5 ideas for LinkedIn posts, each 150 words or fewer,
focused on practical results (not theory).
Format: the idea + a hook for the opening line."

The more specific you are, the better the result. Specificity = Quality.

The three rules of the Contractor Mindset:

1. Context: who you are, who it's for, why
2. Task: exactly what you need
3. Format: what the result should look like

Practice

Exercise: Audit your workday

1. List 5-10 tasks you did over the last 2-3 days
2. For each task, ask: "Could AI have taken on 30% or more of this?"
3. Pick one task from the list
4. Write a specific prompt (a prompt is the text request you give an AI) using the Contractor Mindset, and give it to your assistant right now
5. Compare the result with what you would have ended up with without AI

Goal: find the first work task you'll start handing part of to AI this week.

Common mistakes

❌ Mistake: Handing 100% of a task to AI and not checking the result. ✅ Instead: Use the 30% rule: AI takes the routine part (draft, research, structure). You check, adjust and make the final call. You always stay in control.

❌ Mistake: Giving Claude vague requests like "do something about marketing." ✅ Instead: Use the Contractor Mindset: context (who you are, who it's for) + task (exactly what you need) + format (what the result should look like). The more specific, the better.

❌ Mistake: Deploying code without understanding what it does (Dark Code). ✅ Instead: After every block of code, ask for an explanation: "Explain what this function does in plain English." Write tests. Understand the architecture even if you don't understand every line.

Tools and resources

• Claude.ai: start with simple prompts to get a feel for the Contractor Mindset
• Claude Code: the agent for people who build; you'll need it later, in the first-build module and the library
• Anthropic Prompt Engineering: Anthropic's official guide to prompting
• Obsidian / Notion: note-taking apps for writing down the techniques that work for you

→ Optional, from the library: Agentic workflows vs. traditional automation: how an agent differs from ordinary automation → Optional, from the library: The Four C's framework: a systematic approach to building your own AI system → The previous lesson: How to write a good prompt: the five parts of a good prompt

Key takeaways

The Default Shift isn't about a tool. It's about deciding that your job is now "running a system," not "doing tasks."

The 30% rule: before any task, ask "how can AI take on at least a third of this?" Start small, then scale up.

Dark Code doesn't hurt you right away; it does its damage quietly. Understand what you deploy, even if you don't understand every line.

Next lesson

→ AI ethics and safety: hallucinations, attacks, bias: how not to trust AI blindly