How to Use AI Company
Write commands that get results. No fluff, no ambiguity.
Three Ways to Run
AI Company offers three run modes. Choose the right one for your task.
The standard mode. The CEO plans, the Operator executes (writes text, reports, analysis), and the Reviewer checks the work. Fast and cost-effective.
Creates working HTML apps, games, tools, and single-page websites. The Operator writes real code and you get a live preview.
Everything in Build mode, PLUS an AI Builder that deploys it as a fully hosted website with a live URL. For real products.
Credit System
Each invite code comes with a dollar budget and a maximum number of runs. Your balance is deducted as you use AI Company.
~$0.10
Per Run
~$0.60
Per Build
~$6.10
Per Pro build
Real-time
Cost tracking in logs
When your budget or run limit is reached, you won't be able to start new runs. Each run shows its cost in the activity log so you can track spending.
Tell AI Company what you want delivered, not how to think about it.
AI Company is an AI CEO that plans, executes, and reviews. Your job is to give it a clear mission. Its job is to figure out how to accomplish it. The more specific your request, the better the result.
Key Features
File Attachments
Upload PDFs, images, documents, or data files with your task. The CEO uses them as context — attach a competitor's screenshot and say "Create something similar but better."
Request Modifications
After a run completes, request changes without starting over. Say "Make the tone more formal" or "Add a pricing section" — the CEO will revise the existing work.
Web Search
The Operator searches the internet for real data, current prices, reviews, and facts. Sources are cited with URLs so you can verify everything.
Activity Log
See exactly what each AI agent did — the CEO's plan, the Operator's searches and drafts, the Reviewer's feedback. Full transparency with token counts and costs.
What You SHOULD Specify
1. The Deliverable
What concrete output do you want?
"Create a comparison table of..."
"Build a landing page for..."
"List the top 5 reasons why..."
2. Constraints & Parameters
Numbers, limits, and boundaries.
"...for a team of 10"
"...in the US market"
"...published after 2023"
3. Success Criteria
How will you know if it's right?
"Must cite at least 3 recent studies"
"All claims must have source URLs"
"Focus on B2B SaaS companies"
4. Format (if specific)
Structure your output needs.
"Include a comparison table at the end"
"Use bullet points for pros/cons"
"Present as a ranked list with scores"
What You Should NOT Specify
Don't tell it HOW to think
Bad:
"First, search for X, then analyze Y, then compare Z..."
Good:
"Compare X, Y, and Z with pricing data"
Don't be vague
Bad:
"Tell me about AI"
Good:
"List 5 AI tools for email automation with pricing and user ratings"
Don't ask for opinions without criteria
Bad:
"What's the best CRM?"
Good:
"Best CRM for a 5-person sales team under $100/month with Slack integration"
Don't request impossible verification
Bad:
"Predict next quarter's revenue for Apple"
Good:
"Summarize analyst estimates for Apple Q1 2025 from major financial outlets"
Task Templates
Research Template
Example: "List the top 5 project management tools for remote teams under $20/user/month, with a comparison table including pricing, key features, and user ratings"
Analysis Template
Example: "Compare Notion vs Coda for a 20-person startup considering pricing, integrations, learning curve, and real user reviews from G2"
Content Template
Example: "Write a 400-word executive summary about remote work trends in 2025 for C-suite executives, citing recent studies"
Website Template
Example: "Create a landing page for a fitness studio with hero section, class schedule, pricing plans, and testimonials using a bold dark theme with orange accents"
Ready-to-Use Commands
Copy any of these complete prompts and paste them directly into AI Company. Each one is a real-world task with recommended settings — just click copy and run.
What happens: The CEO will plan a research strategy, the Operator will conduct multiple web searches to gather data from scientific journals, company press releases, and news articles, and the Reviewer will verify that all claims are properly sourced and accurate.
What happens: The CEO will analyze your constraints and skills, the Operator will research real business models and success stories, and the Reviewer will challenge the feasibility of each idea to ensure the final ranking is realistic, not just optimistic.
What happens: The CEO will plan the website structure and design approach, the Operator will build the complete HTML/CSS page with all requested sections, and the Reviewer will check for design consistency, responsiveness, and content accuracy. You'll get a live preview you can download.
What happens: The CEO will structure the comparison framework, the Operator will research each tool across G2, Capterra, and official pricing pages, and the Reviewer will verify that all ratings, prices, and feature claims are accurate and properly sourced.
Quick Reference
| Task Type | Mode | Cycles | Searches | Budget |
|---|---|---|---|---|
| Simple question / math | Run | 1 | 0-2 | $0.50 |
| Quick research / fact check | Run | 2 | 3-5 | $1-2 |
| Website / landing page | Build | 3-4 | 3-8 | $2-4 |
| Comparison / analysis | Run | 3-5 | 5-12 | $2-5 |
| Deep research / report | Run | 5-7 | 10-15 | $4-7 |
| Deployed multi-page site | Pro | 3-4 | 3-8 | $3-7 |
| Full web application | Pro | 5-8 | 5-15 | $5-13 |
Pro Tips
Use "Request Modifications" after a run
After the CEO delivers a draft, you can request modifications without starting over. Tell it "Make the tone more formal" or "Add a section about pricing" — the CEO will revise the existing work. You can also attach images or files for reference.
Check the Activity Log
The Log tab shows exactly what each AI agent did — what the CEO planned, what the Operator searched and wrote, and what the Reviewer flagged. Use it to understand the process and refine your next command.
Budget controls prevent runaway costs
Set a Max Dollar amount to cap spending. The run will stop when the budget is reached, even if cycles remain. For exploration, start with $2-3 and increase if you need deeper analysis. Set "E" (endless) for cycles/searches if you want the budget to be the only limit.
Attach files for context
Upload PDFs, images, documents, or data files with your task brief. The CEO will use them as context — for example, attach a competitor's website screenshot and say "Create something similar but better for my brand."
Start with "Run" mode, upgrade to "Run + Build" when needed
Most tasks work great with the standard "Run" button (Operator mode). Only use "Run + Build" when you specifically need a deployed website or application. Builder mode costs more and takes longer, so save it for when you need the real thing.