Building with AI, Your First AI Project

You're Ready to Build

You've put in the work. Five modules in, you understand what AI is, how it works, where it fits, and where it falls short. That's more than most people will ever know about this stuff.

Now it's time to do something with it.

This module is about moving from understanding to creating. Not in a "start a company" way or a "learn to code" way. In a practical, this-weekend, right-now way. By the end, you'll have built at least one real AI tool that saves you time -- and you'll have a process for building more.

The gap between "I understand AI" and "I use AI to build things" is smaller than you think. Let's close it.

What "Building with AI" Actually Means

"Building" sounds intimidating. It conjures images of developers hunched over keyboards, debugging code at 2 a.m., drinking cold coffee.

Forget that image.

Building with AI means using AI tools to create something that solves a problem. That's it. The something might be a custom assistant that writes like you. It might be a workflow that turns rough notes into polished emails. It might be a system that researches a topic and hands you a summary.

You're not writing software. You're not starting a business. You're combining existing AI tools in a way that works for you.

Think of it like cooking. You're not inventing a new cuisine. You're following a recipe, then adjusting it to your taste. The ingredients are AI tools. The recipe is a process. The meal is whatever saves you time or makes you money.

There are three levels of building with AI, and they build on each other naturally:

Level 1: Personal Tool -- Something that saves you time. A custom AI assistant that knows your preferences. A workflow that automates a task you do every week. This level is about you. Build it for yourself. Use it yourself. If it saves you 30 minutes a day, that's a win.

Level 2: Shared Tool -- Something that helps others. You take what works for you and adapt it so a few other people can use it. Maybe you share your custom GPT with a colleague. Maybe you turn your research workflow into a template. This level is about proving your tool works beyond just you.

Level 3: Sellable Tool -- Something that earns money. Your tool is so useful that people would pay for it. You package it, price it, and offer it. This level is rare, and that's fine. Most valuable AI tools stay at Level 1 or 2 forever, and that's perfectly okay.

The important thing: you don't start at Level 3. You start at Level 1. Some tools never move past Level 1, and they're still worth building. A tool that saves you two hours a week is worth two hours a week. That's the point.

Five AI Projects Anyone Can Build This Weekend

These projects are designed to be built in a single weekend with free tools. No coding. No prior experience beyond what you've learned in this course. Each one teaches you a different skill, so even if you only build one, you'll come away knowing more than when you started.

Project 1: Personal AI Assistant

What it does: A custom AI assistant that knows your writing style, your preferences, your common tasks, and your context. Instead of explaining yourself every time you open ChatGPT or Claude, your assistant already knows who you are and what you need.

Tools needed: ChatGPT Plus (Custom GPT) or Claude Pro (Project) -- both offer free trials if you don't have one yet.

Steps to build:

Open ChatGPT and click "Explore GPTs," then "Create a GPT." Or open Claude and start a new Project.
Write a system prompt (instructions) that tells the AI who you are. Include: your job, your writing tone, your common tasks, things you always want and things you never want. Be specific. "I'm a freelance marketer who writes in a conversational tone. Never use corporate jargon. Always suggest specific examples, not vague advice."
Add your commonly used documents as reference files -- style guides, past writing samples, templates you reuse.
Test it with three real tasks you do regularly. Ask it to write an email, draft a social post, outline a blog post. See if it sounds like you.
Adjust the instructions based on what misses the mark. If it's too formal, add "Write like you're talking to a friend at a coffee shop." If it's too generic, add "Always include a specific example or anecdote."
Use it for a week. Tweak as you go.

Time required: 30 minutes to set up. 15 minutes per day for a week to refine.

What you'll learn: How to write effective system prompts (the most underrated AI skill). How to give AI context so it works better for you. The difference between a generic AI response and a personalized one.

Project 2: AI Newsletter

What it does: A system that researches a topic, drafts a newsletter issue, and formats it for email -- all guided by you. You provide the topic and editorial direction. AI does the heavy lifting of research and drafting. You edit and send.

Tools needed: ChatGPT or Claude (free tier works), a document editor (Google Docs, Notion, or similar), your email platform (Substack, Beehiiv, Mailchimp -- all have free tiers).

Steps to build:

Pick your newsletter topic. Choose something you already follow and care about -- AI tools for small business, local restaurant news, personal finance tips for freelancers. The more specific, the better.
Create a template for your newsletter. Write out the sections: introduction, main story, quick links, sign-off. Keep it simple. Three to five sections max.
Build a research prompt. Tell AI: "I'm writing a newsletter about [topic]. Give me the 5 most important developments this week, with a one-paragraph summary of each. Include sources."
Build a drafting prompt. Tell AI: "Using these research notes, draft a newsletter issue. Follow this format: [paste your template]. Write in a [describe your tone] voice. Keep the total length under 600 words."
Run the workflow: research prompt first, review the output, then feed it into the drafting prompt.
Edit the draft. Add your opinions, cut what's boring, fix what sounds robotic. This step is non-negotiable -- AI drafts are starting points, not finished products.
Copy into your email platform, format, and send.

Time required: 1 hour to set up templates and prompts. 30-60 minutes per issue once the system is running.

What you'll learn: How to break a complex task into smaller AI prompts. How to chain prompts together (research, then draft, then edit). How to add your voice to AI-generated content. The discipline of editing AI output instead of publishing it raw.

Project 3: AI Content Machine

What it does: Takes one set of ideas or notes and turns them into multiple pieces of content -- social media posts, blog outlines, email drafts, and talking points. You think once, publish many times.

Tools needed: ChatGPT or Claude (free tier works), a spreadsheet or document to track outputs.

Steps to build:

Create a "content input" template. This is where you write your core idea. Example: "Three reasons most small businesses waste money on ads: they don't track ROI, they copy big-brand strategies, they quit too early."
Build a social media prompt: "Turn this idea into 5 social media posts. One for LinkedIn (professional, insightful), one for Twitter/X (punchy, contrarian), one for Instagram (story-driven, visual prompt), one for a Facebook group (conversational, asking for input), and one short-form video script (30 seconds, hook-driven)."
Build a blog outline prompt: "Turn this idea into a blog post outline. Include: headline options (3), introduction hook, 4-5 main sections with key points, and a conclusion with a call to action."
Build an email prompt: "Turn this idea into a short email to my subscribers. Casual tone. One main point. End with a question to spark replies."
Build a talking points prompt: "Turn this idea into 5 bullet-point talking points I can use in a meeting, podcast, or presentation."
Run all four prompts with the same input. Review each output. Edit for your voice and accuracy.
Schedule or publish the content across your channels.

Time required: 45 minutes to build prompts. 20-30 minutes per content batch.

What you'll learn: How to repurpose one idea across multiple formats. How to write prompts that adapt output for different platforms. How to create a repeatable content system instead of starting from scratch every time.

Project 4: AI Research Tool

What it does: A workflow that takes a topic, researches it, and produces a formatted report with sources. You go from "I need to understand X" to "Here's a clear summary with citations" in minutes instead of hours.

Tools needed: ChatGPT or Claude (free tier), a document to store reports, optionally Perplexity AI (free tier) for source-backed research.

Steps to build:

Define what a "good report" looks like for you. One page? Three pages? Executive summary at the top? Bullet points or paragraphs? Write this down -- it becomes your format template.
Create a research prompt: "Research [topic]. Focus on the most current information available. For each key finding, provide: a one-sentence summary, a paragraph of detail, and the source. Cover at least 5 key findings."
Create a synthesis prompt: "Take these research notes and turn them into a formatted report. Structure: executive summary (3-4 sentences), key findings (bulleted), detailed analysis (2-3 paragraphs), and sources list. Write in clear, plain language. Avoid jargon."
Run the research prompt first. Review for accuracy and gaps. Add follow-up questions if needed: "I need more detail on finding #3" or "Are there any opposing viewpoints?"
Feed the complete research into the synthesis prompt. Review the output.
Always verify key claims yourself. AI research is a head start, not a replacement for checking facts. Pick at least two claims and confirm them with original sources.
Save the report in your preferred format. Build a library over time.

Time required: 45 minutes to build prompts and templates. 15-30 minutes per research report.

What you'll learn: How to use AI for research without trusting it blindly. How to chain research and synthesis prompts. How to verify AI-sourced information. The habit of checking, not just accepting.

Project 5: AI Customer Reply System

What it does: A system that drafts customer responses in seconds instead of minutes. You paste the customer's message, and AI generates a reply that matches your brand voice and addresses their concern. You review, tweak, and send.

Tools needed: ChatGPT or Claude (free tier), a document with your customer response policies and common scenarios.

Steps to build:

List your 10 most common customer inquiries. Returns, shipping questions, billing issues, feature requests, complaints -- whatever comes up most often.
For each, write your ideal response. Not a template with blanks -- a real response you'd be proud to send. This is your training data.
Write a brand voice guide. Three to five sentences describing how your company sounds. Example: "Friendly but professional. We apologize first, then solve. We use the customer's name. We never blame the customer. We always offer a next step."
Create a master prompt: "A customer sent this message: [paste message]. Draft a reply that follows these guidelines: [paste brand voice guide]. Here are examples of how we respond to similar situations: [paste your example responses]. Keep the reply under 150 words. Address their specific concern. Offer a clear next step."
Test with 5 real customer messages from your inbox. Compare AI drafts to what you actually sent. Adjust the prompt until the drafts are close to your real responses.
Build the system into your workflow: when a customer message comes in, paste it into your AI tool, review the draft, edit as needed, and send.
Track how much time you save per response. Most people cut reply time by 60-80%.

Time required: 1 hour to build. 5-10 seconds per draft response.

What you'll learn: How to give AI examples to learn from (few-shot prompting). How to encode your brand voice into a prompt. How to build trust in AI output gradually by comparing it to your own work.

The Build Process (Works for Any Project)

Every project in this module follows the same process. It's simple on purpose. Simple processes get finished. Complex processes get abandoned.

Step 1: Start with a problem you have.

Not a problem you imagine. Not a problem someone else has. A problem you personally experience. "I spend two hours a week writing the same type of emails." "I can never find good research on topics I need." "I start content but never finish because the drafting takes too long."

Your problem is your project. If you don't have the problem, you won't have the motivation to finish the project.

Step 2: Sketch the solution on paper.

Draw it. Literally on paper. Three boxes: inputs, AI processing, outputs.

Inputs: What information goes in? (A customer message. A topic. A set of notes.) AI processing: What does AI do with it? (Draft a reply. Research and summarize. Turn notes into content.) Outputs: What comes out? (A polished email. A research report. Five social media posts.)

This sketch takes five minutes and saves you from building the wrong thing.

Step 3: Build a rough version in one hour.

Not a perfect version. A rough version. Write the prompt. Test it once. See what happens. If the output is terrible, adjust the prompt. If it's okay, move on. The goal is a working prototype, not a finished product.

Set a timer. One hour. When it goes off, you have version 1. It will be imperfect. That's the point.

Step 4: Test it on five real examples.

Not hypothetical examples. Real ones. Real customer messages. Real topics you need researched. Real content ideas you want to develop. Five is enough to see patterns -- what works, what doesn't, what needs adjusting.

Step 5: Fix what breaks.

After five tests, you'll know what's wrong. The output is too long. The tone is off. It misses a key piece of information. Fix the prompt. Test again. Repeat until the output is good enough to use -- not perfect, good enough.

Perfection is the enemy of finished. Ship version 1. Improve it later.

Step 6: Use it for a week before adding features.

The temptation after building something is to immediately make it bigger. Add more prompts. Handle more cases. Automate more steps. Resist this.

Use your tool for a week. Notice what actually breaks in real use, not what you imagine might break. Then fix only what actually breaks. This is how you build something that works instead of something that's impressive on paper but falls apart in practice.

No-Code Tools for Building

You don't need to code to build with AI. The tools below are all you need, and they all have free options.

ChatGPT Custom GPTs (requires ChatGPT Plus at $20/month) -- Custom GPTs let you create a personalized version of ChatGPT with your own instructions, knowledge files, and capabilities. You build it once, and it's available every time you open ChatGPT. Best for: assistants, content tools, and anything that benefits from persistent context.

Claude Projects (requires Claude Pro at $20/month) -- Projects let you upload documents and set instructions that Claude remembers across conversations within the project. Similar to Custom GPTs but with a stronger focus on document analysis and long-form writing. Best for: research tools, writing assistants, and projects that need reference documents.

Make.com (free tier available) -- A visual workflow builder that connects different apps and AI services. You drag and drop modules to create automations: "When I get an email, send it to AI, then save the draft to Google Docs." Free tier includes 1,000 operations per month. Best for: multi-step workflows, connecting AI to other tools, and automating repetitive processes.

Notion AI (free tier available, AI features have a small add-on cost) -- Notion's built-in AI can write, summarize, brainstorm, and translate within your Notion workspace. Best for: document-based tools, knowledge management, and projects where your notes and AI output live side by side.

Google Sheets + AI (free with Google account) -- Connect Google Sheets to AI using simple add-ons like GPT for Sheets or via Make.com. Best for: data tools, batch processing, and any project where you need AI to work through rows of information.

You don't need all of these. Pick one or two that match the project you want to build. Most people start with ChatGPT or Claude and add Make.com when they want to connect things together.

From Personal Tool to Something Bigger

You built something. It works. It saves you time. Now what?

Most AI tools stay personal, and that's fine. But some have the potential to become more. Here's how to tell the difference and what to do about it.

When your tool saves you 5+ hours per week, share it with 3 people.

Not the whole internet. Three people. Colleagues, friends, or someone in your industry who has the same problem. Share it simply: "I built this thing that helps me do X. Want to try it?"

Watch how they use it. Do they come back and use it again? Do they ask questions about how it works? Do they suggest improvements? If yes, you might have something.

When 3 people love it, consider making it a product.

"Love" means they use it regularly, not that they said "neat" once. If three people are genuinely using your tool and finding it valuable, it's worth thinking about packaging it for a wider audience.

This doesn't mean quit your job and start a company. It means: clean up the instructions, write a short guide, and offer it to 10 more people. Maybe charge a small amount. See what happens.

When strangers ask for it, you might have a business.

The strongest signal isn't people you shared it with liking it. It's people you don't know asking for it. When someone finds your tool through word of mouth or a post and says "I need this," that's market demand. That's when it's worth investing real time into turning a tool into a product.

But here's the honest truth: most tools don't reach this level. And that's okay. A tool that saves you time is valuable even if no one else ever uses it. Don't skip building because you're not sure it'll become a business. Build because it helps you. Everything after that is a bonus.

Try It: Build Your Personal AI Assistant

This is the fastest, highest-impact project in the module. Fifteen minutes to set up. Saves you hours per week. Let's build it right now.

Option A: Build a Custom GPT (ChatGPT Plus)

Open ChatGPT. Click "Explore GPTs" in the sidebar, then click "Create" at the top right.
Name your GPT something clear: "My Assistant" or "Jimmy's Helper" or whatever feels right.
In the Instructions field, write the following (customize the bracketed parts to fit you):

You are my personal assistant. Here's what you need to know about me:

- My name: [your name]
- My job: [what you do]
- My writing style: [conversational / professional / casual / technical]
- Things I always want: [specific examples, action items, short answers]
- Things I never want: [jargon, vague advice, responses over 300 words unless I ask]

My common tasks:
1. [Task you do often, e.g., "Draft emails to clients"]
2. [Another task, e.g., "Summarize long articles"]
3. [Another task, e.g., "Brainstorm content ideas"]

When I ask you to write something, match my writing style.
When I ask for advice, be direct. No hedging.
When I ask you to summarize, keep it under 5 bullet points.
Always ask me one clarifying question before starting a complex task.

Under "Knowledge," upload any documents that help the AI understand your work: past writing samples, a style guide, templates you reuse, your product descriptions. Up to 20 files.
Under "Capabilities," turn on Web Browsing if you want the assistant to search the internet. Turn on Image Generation if that's useful to you. Leave Code Interpreter on -- it's helpful for data tasks.
Save it. Test it with a real task. Adjust the instructions based on the output.

Option B: Build a Claude Project (Claude Pro)

Open Claude. Click "Projects" in the sidebar, then click "Create Project."
Name your project.
In the Project Instructions field, write the same instructions from Step 3 above.
Upload your reference documents as project files.
Start a new conversation within the project. The instructions and files are automatically available to Claude in every conversation within this project.
Test with a real task. Adjust instructions as needed.

Making It Yours

The instructions above are a starting point. After a week of use, you'll know exactly what to add:

Phrases the AI uses that annoy you? Add them to "Things I never want."
Types of tasks that come up repeatedly? Add them to "My common tasks."
Formats you always want? Add them: "Always format emails with a greeting, two short paragraphs, and a sign-off."
Context about your audience? Add it: "My clients are small business owners who hate corporate speak."

The magic of a personal AI assistant isn't that it's smarter than regular AI. It's that it already knows you. Every tweak you make to the instructions is an investment that pays off every time you use it.

Fifteen minutes of setup. Hours saved per week. That's the best return on time you'll find anywhere in this course.

Key Takeaways from Module 6

Building with AI means creating something useful, not coding. You combine existing AI tools to solve a problem you have. That's building.
Start at Level 1: build for yourself. A tool that saves you time is worth building even if no one else ever uses it.
Five projects, one weekend. Personal assistant, newsletter, content machine, research tool, customer reply system -- pick one and start.
The build process is the same every time. Problem, sketch, rough version, five real tests, fix what breaks, use it for a week. Six steps.
No-code tools are enough. ChatGPT, Claude, Make.com, Notion, Google Sheets. You don't need to code. You need to be clear about what you want.
Share when it works, productize when strangers ask. Don't build for a market. Build for yourself. Let demand find you.
Your personal AI assistant is the highest-return project. Fifteen minutes to set up. Hours saved per week. Build it today.
Done beats perfect. Build a rough version in one hour. Use it. Improve it. Don't plan for weeks before building anything.

What's Next

You've gone from understanding AI to building with it. That's a big step. But building something powerful without understanding the risks is like driving a fast car without knowing where the brakes are.

Module 7 -- "AI Safety, Ethics, and What Can Go Wrong" -- covers the stuff most AI courses skip or breeze through. We'll look at what happens when AI systems make mistakes, who's responsible when they do, how bias creeps into AI outputs, and what you can do to use AI responsibly without becoming paranoid about it.

Because the goal isn't to avoid AI's problems. It's to understand them well enough to build anyway.