Start a 1-Person Business with Claude (4 HOUR COURSE 2026)

Overview

Albert Olgaard delivers a 4-hour course on building a one-person AI business using Claude Code, covering client acquisition through Upwork and cold email via Instantly.ai, sales call structure, and service delivery for websites, automations, and full AI systems. He demonstrates building production apps using Next.js, Trigger.dev, Composio, and MongoDB, while teaching a framework of build plans and implementation plans for tackling any project, plus pricing strategies based on a 5x ROI rule.

Key takeaways

The biggest opportunity in 2026 isn't being early to AI — it's that 84% of 8.1 billion people have never used it, and only 0.04% (3.6M people) use advanced models like Claude Code, creating a massive service gap.
Upwork's competition is weaker than perceived because most freelancers come from Pakistan, India, and Nigeria using templated AI-generated proposals — personalized cover letters with Loom videos and Western-country profiles stand out instantly.
Cold email reply rates in unique-language markets (Swedish, Portuguese, Greek) hit 5%+ because of dramatically lower competition, compared to 0.5% typical in English markets.
The 5x ROI pricing rule (charge 20% of value created) keeps churn near zero — Shiny.ai's result-based deals produced one client worth $240K LTV that would have been $10K on a flat retainer.
Trigger.dev's open-source nature lets agencies build automations once and deploy them anywhere (Vercel, AWS, client's own servers), unlike n8n/Make which lock you into proprietary platforms.
The proof-of-concept satisfaction guarantee (full payment upfront, full refund if client dislikes the build) closes more deals because it transfers all risk to the agency — Albert has used it dozens of times without a single refund.
Building Claude skills from completed projects is the agency equivalent of SOPs — the real long-term value isn't the code but the accumulated process documentation that makes each new client onboarding faster.

Chapters

0:00 Why AI Creates an Unprecedented Opportunity for Solo Founders

Albert built two AI companies generating over $1M combined with no CS degree
Statistical breakdown: 84% of 8.1B humans have never used AI, only 0.04% (3.6M people) use advanced models like Claude Code
The gap between AI power users and the general population creates the service opportunity

5:00 The Excitement Curve and Why 95% of Beginners Quit

New business excitement spikes then crashes when reality hits — landing first clients is hard
Top 5% who push through the dip eventually crack the business model
Warning against jumping between business models (dropshipping, NFTs, crypto) instead of committing

8:00 Course Roadmap: Clients First, Then Build

Order matters: start with client acquisition (Upwork + cold email) before learning tech
Covers booking meetings, using Claude to listen to calls via Fathom MCP
Service delivery for websites, automations, mini apps, then automating your own business

10:00 Setting Up Claude Code in Visual Studio Code

Install VS Code, create a project folder (named 'shiny' for the agency), open terminal
Run install command from claude code docs, authenticate with Claude Pro or Max subscription ($20/month minimum recommended)
Pro plan recommended for beginners over API key approach

14:10 Installing Claude Skills for Enhanced Capabilities

Download Albert's skills bundle from the free 1% in AI community drive folder
Use Claude to research documentation and install skills in the project's .claude folder
Restart session with Ctrl+C and 'claude' command to enable new skills like /front-end-design

17:00 Backing Up Your Workspace to GitHub

Create a private GitHub repo (e.g., 'shiny workspace') to preserve skills and configs
Use Claude to push the project folder to the cloud repository
Set to default (not auto) mode to allow Claude to execute git push commands

19:00 The 'Get Good Before Get Rich' Principle

Albert's first paying client took 4 months and paid only $400
Trust requires testimonials and previous experience — start with free work to build both
Reject the Instagram fantasy of selling $10K websites as a beginner

22:10 Three Service Levels: Websites, Automations, AI Systems

Level 1: AI-generated websites — 30% of US small businesses (10M+) still don't have one
Level 2: Automations and agents requiring API and integration knowledge
Level 3: Full AI systems combining multiple automations with dashboards

26:00 Joining the 1% in AI Community for Claude Code Mastery

Albert personally answers every post in the community
30-day AI challenge refunds your first month if you complete it consistently
Includes full Claude Code masterclass for beginners

27:00 Why Start with Upwork: First Clients Are the Hardest

Upwork connects freelancers with businesses needing AI services directly
Search for specific niches like Go High Level, n8n, or AI automation
Levels to the game — profile setup and applications matter significantly

29:00 Creating a High-Converting Upwork Profile

Import from LinkedIn PDF or fill manually, choose IT/CRM categories
Add skills like Go High Level, n8n, Make, Sabia (Zapier), Claude, TypeScript, Python
Set rate at $15/hour to start — perfect score matters more than hourly rate for first jobs

34:00 Building Trust Signals on Your Profile

Complete the working style assessment to earn 'Clear Communicator' badge
Add portfolio with real screenshots like booked Google Calendars and AI agent dashboards
Set location to a recognizable city (Copenhagen, not a small town) for credibility

36:00 Studying Million-Dollar Upwork Profiles

Top earners like Vipa D ($1M earned) share patterns: video consultation offers, top-rated badges
Steal structures from successful profiles — they're 100% public
98-100% job success rate is the single most important profile metric

38:00 Using Claude's Upwork Skill to Optimize Your Profile

The /upwork skill is trained on top-performing profiles in the AI agency space
Albert rewrites his title to 'AI Automation Engineer | AI Agents & GoHighLevel | Make & n8n'
Use precise pricing like $16.73/hour instead of $16 to signal calculated value

40:00 Upgrading to Upwork Freelancer Plus

50% off first month ($10) gives 100 connects vs. 10 on free plan
Verify identity (costs 35 connects) and enable 'Available Now' badge (14 connects/week)
Both significantly increase trust and visibility

42:20 Bidding Strategy: Speed and Filtering

Avoid jobs with 50+ proposals — target the 5-20 proposal range
Be instant when new jobs are posted — speed-to-lead matters
Verify client legitimacy with payment-verified and previous spend stats

45:00 Writing Proposals with Claude's Proposal Skill

The /proposal skill flags red-flag clients (e.g., $10 fixed price with 67 open jobs = hiring fishing)
Educate clients in proposals by mentioning specific technical insights (e.g., Go High Level v2 API)
Add 'this is not written by AI' framing and personalize manually

49:00 The Power of Adding a Loom Video to Every Proposal

Show your face on camera — proves you're a real person, not AI
Demonstrate specific knowledge like the Add Account API call for auto-deploying sub-accounts
Even a 1-minute loom dramatically increases close rate

52:00 Why Western Freelancers Have an Edge on Upwork

Most Upwork competition comes from Pakistan, Nigeria, India with templated AI-generated proposals
Personalized cover letters written like humans stand out instantly
Building Upwork takes months but compounds — clients refer to other clients

57:00 Cold Email with Instantly.ai for Higher Volume

Map of countries shows blue regions (Europe, South America, Asia) are emerging AI markets
Cold email works exceptionally well in blue countries due to lower competition
Instantly.ai has an MCP allowing Claude to do 90% of campaign work

59:00 Connecting Instantly's MCP to Claude Desktop

Download Claude desktop app, add custom connector with Instantly API key
Verify connection with /mcp command — should show 38 available tools
Always install MCPs in Claude Desktop first, they propagate to Claude Code

1:02:00 Buying Pre-Warmed Email Domains in Instantly

Pre-warmed accounts cost ~$15/month per domain, each sends 20 emails/day
Choose professional-sounding domains (e.g., land-ai-magine.org)
Set forwarding domain to your real website (e.g., shiny.ai)

1:05:00 Building Cold Email Campaigns with the Instantly Skill

Real campaign example: 1,000 emails to dentists, 5.1% reply rate, 50 replies, 6 interested leads
Target unique-language markets (Swedish, Greek, Portuguese) where competition is lower
Use 'lost revenue calculator' as the lead magnet/subject line hook

1:09:00 Scraping Leads with Apollo + TrustedLeads.io

Use Apollo's filters: location, industry, job titles (CEO/Founder/Owner), company size (1-50 employees)
Always enable 'verified emails' filter for quality
TrustedLeads.io scrapes Apollo URLs at $0.005/lead — 67% cheaper than Apollo direct

1:13:00 Verifying Emails with MillionVerifier Before Sending

Run all leads through MillionVerifier — typical result: 64% good, 17% risky, 18% bad
Only send to 'good' emails to protect domain reputation
Bouncing emails get domains blacklisted instantly

1:16:00 Launching the Cold Email Campaign

Map CSV fields to Instantly (first name, last name, company, website, email)
Add pre-warmed accounts to campaign options before launching
Send during business hours (7am-4pm), avoid weekends

1:18:00 Sales Call Rule #1: Never Mention Price Before the Call

Pre-call price quotes get judged on price, not value
Hopping on a call lets you show $10K of value before quoting $2K
Always send confirmation + reminders on multiple channels — at booking, 3 days, 1 day, 1 hour, 5 minutes before

1:22:00 The Doctor Approach: 80/20 Talk Ratio on Calls

Client talks 80% of time, you talk 20% — mostly questions
Get clients to verbalize their own pain (e.g., 'competitor has 150 reviews, you have 5')
Self-stated problems convert dramatically better than salesperson-stated problems

1:24:00 Call Structure: Pain, Solution, Price, Objections

Step 1: Questions positioning a pain — Step 2: Present solution — Step 3: Reveal price last
Never use a slideshow that just lists features — that's a beginner mistake
Handle objections at the end, then close

1:27:00 Always Book the Next Call on the Current Call

Calendar invites massively reduce ghosting
Hype the next call by saying you'll have specific work prepared
Make clients feel invested before they leave

1:28:20 Take Payment Commitment Even for Free Trials

Create a $1/year Stripe payment link to save payment info on file
Removes friction when converting trials to paid
Even free trials should require credit card commitment

1:30:00 The Proof-of-Concept Satisfaction Guarantee Offer

Take full payment upfront ($1,000+) but offer full refund if client dislikes the build
Removes all client risk and signals confidence
Albert has used this dozens of times and never had to refund — clients always commit after seeing the work

1:32:00 Post-Sale Process: Bi-Weekly Calls, Upsells, Referrals

Bi-weekly calls reduce churn from 20-30% down to ~5%
Specific referral asks ('Do you know any plumbers or electricians?') beat generic ones
Offer 20% commission to incentivize referrals

1:36:20 Using Fathom MCP to Auto-Capture Call Context

Install Fathom (or any AI noteaker) and connect via MCP browse connectors
Claude can pull meeting summaries and transcripts directly after calls
Eliminates losing context between sales call and service delivery

1:39:00 The Web Agency Pricing Reality Check

Solo freelancer template sites: $1-3K; small boutique: $3-10K; proper agency: $10-30K
Mid-tier agencies charge $30-80K, top-tier $80-250K, brand-name agencies $250K+
Albert's agency recommends ~$10K for websites once skilled

1:44:00 Building Premium Websites with the Build-Premium-Website Skill

The /build-premium-website skill produced a full landscaping site (Cityscape) in 8 minutes
Includes animated hero, services grid, process steps, contact form — fully responsive on mobile
Use Mars green color palette for bold modern themes

1:49:00 Deploying Websites with GitHub + Vercel

Push code to public GitHub repo, then import into Vercel
Vercel auto-detects Next.js and deploys in ~9 seconds with free hosting tier
Add custom domains in Vercel's domains tab — connect via DNS

1:52:00 Automations with Trigger.dev Instead of n8n/Make

Trigger.dev is open-source (15K GitHub stars), allowing self-hosting on client servers
Claude codes Trigger.dev workflows natively — no visual builder needed
Use the /trigger.dev skill combined with the /composio skill for auth

1:54:00 Building an Automated Invoicing System

Form on Next.js front-end captures business info, line items, tax/discounts
Trigger.dev generates PDF with React, attaches to Gmail draft, stores in Google Drive
Composio handles Gmail and Drive OAuth via shareable URLs

1:59:00 Configuring Environment Variables and OAuth Flow

Fill .env.local with Trigger secret key, Composio API key, Anthropic API key, Google Drive folder ID
Composio prints OAuth URLs to terminal — click each to authorize Gmail and Drive
AI-generated email subjects and bodies use Anthropic API

2:05:00 Deploying Automations to Trigger.dev Production

Set production environment variables in Trigger.dev dashboard
Push tasks via Claude with the production project ref
Dashboard shows execution logs, errors, and run history

2:10:00 Creating Reusable Skills from Built Projects

Use the /create-skill skill to generalize patterns from completed builds
Built a generic 'mini-automation' skill from the invoicing project
Skills become company IP — the equivalent of SOPs in traditional businesses

2:14:00 Building Full AI Systems: Architecture Overview

Stack: Next.js + Tailwind + NextAuth + Resend (front-end), Trigger.dev (back-end), MongoDB (database)
Magic link authentication is more secure than passwords — forces 2FA every login
Domain-restricted login (e.g., only @shiny.ai emails) adds enterprise-grade security

2:18:00 Scaffolding the AI System Dashboard

Claude builds login, sidebar, dashboard cards, settings page in one prompt with plan mode
Use front-end-design skill afterward for Linear-style light/dark mode makeover
MongoDB stores users, automation runs, and embeddings in one cluster

2:24:00 Installing the First Tool: Invoice Automation Inside the Dashboard

Reuse Trigger.dev backend logic, port front-end form into the dashboard
Add OAuth modal flow for client-side Composio authentication
Each automation becomes a sidebar item users can trigger

2:29:00 Adding an Onboarding Automation

Sends contract PDF, Calendly link, and onboarding survey via Gmail composio
Generates draft PDF contract dynamically with client name and business info
Built in ~3 minutes since the system architecture already existed

2:35:00 Building an AI Support Email Agent with RAG

Polls Gmail every 10 minutes for new emails, AI replies to ~60% with high confidence
Knowledge base uses MongoDB vector search for embeddings — no separate Pinecone needed
Escalates low-confidence emails to a human-review tab with AI-pre-drafted replies

2:42:00 Testing the AI Support Agent End-to-End

Albert sent a real test email — AI replied within minutes citing knowledge base sources
Each reply shows confidence score and source citations for transparency
Could replace 5 of 10 support reps saving $10K/month, justifying $2-5K/month pricing

2:51:00 Deploying the Front-End AI System to Vercel

Push only the front-end folder to a separate GitHub repo
Fix MongoDB peer dependency conflict with --legacy-peer-deps
Critical: update NEXTAUTH_URL environment variable to the production Vercel URL

2:57:00 The Four-Step Framework for Building Anything

Step 1: Build plan (tech stack, design, features) — Step 2: Implementation plan with checkbox subtasks
Step 3: Execute step by step — Step 4: Test and refine bugs
Spend 80% of time on steps 1-2, only 20% on actual coding

3:03:00 Live Demo: Building a Lead Enricher from Scratch

Uses Hunter.io API as primary, supports email/domain/name/company inputs
Anthropic's Superpowers plugin (28K GitHub stars) forces engineering-grade planning
Found Albert's co-founder Oliver's email from just 'shiny.ai' domain

3:13:00 Pricing Rule: The 5x Return on Investment Standard

Save client $10K/month → charge $2K/month maximum
Keeps churn low and generates referrals, potentially achieving negative churn
Calculate ROI from labor savings (e.g., 5 support reps × $2K) or new revenue × gross margin

3:17:00 Pricing Models: Result-Based vs. Retainer vs. Hybrid

Result-based (20% of gross profit) works for big clients with trackable revenue — one client paid Shiny.ai $240K LTV
Upfront + recurring with satisfaction guarantee removes risk while ensuring commitment
Use tiered pricing for usage-based services to protect margins

3:22:00 Software Costs and Selling Outcomes Not Features

Include software costs in your price for simplicity — confused buyers don't buy
Only break out costs for enterprise clients hosting on their own servers
Clients buy outcomes ($10K saved), not AI features or model names

3:27:00 Five-Step Roadmap from Beginner to Scaling

Step 1: Learn the skill — Step 2: Land first clients (free or cheap) — Step 3: Sustain yourself
Step 4: Charge more, work with fewer better clients — Step 5: Find and fix bottlenecks indefinitely
Spend 4 hours daily on marketing/sales when starting — not just learning AI

3:33:00 Building the Business Around Your Day Job

Use the 5-9 before 9-5 pattern: 4 hours morning + 2 hours evening + 10 weekend hours
Don't quit your job until the business sustains you — this is a myth that kills starters
Albert's first $400 client took 4 months of daily grinding

Summary, takeaways, and chapters were generated by AI from the video's transcript and may contain errors. The video belongs to its creator, Albert Olgaard.