Quick Start

Quick Start: From Idea to Artifacts in 15 Minutes

This section is a step-by-step guide for your first time using AI CPO. Follow the steps in order, and within 15-20 minutes you'll have your first set of product artifacts.

Step 1: Open the Platform

Go to the home page. You can start without registration — the platform creates a "shadow" project tied to your session. All data is stored in the database and won't be lost. When you register, the project automatically binds to your account.

Tip

Register right away if you plan to work with multiple projects or return later. Shadow projects are tied to the browser session.

Step 2: Describe Your Product

Type a description of your product, idea, or problem in the chat. The more detail in your first message, the faster AI will collect context and unlock artifacts. Try to cover all 5 dimensions:

Dimension	What to Write
Niche	What industry? What market?
Pains	What problem are you solving? How does it manifest?
Audience	Who is your customer? What situation are they in?
Economics	How much do alternatives cost? What is the customer willing to pay?
Competitors	What do people use now? What are they unhappy with?

Good First Message Example

"I want to build a mobile task management app for freelance designers. The core problem is they waste time switching between 5-6 tools (Figma, Notion, Trello, messengers). Competitors are Notion and ClickUp, but they're too complex for a solo worker. Target audience: freelancers earning $1,000-3,000/month managing 3-5 projects simultaneously. Toggl costs $10/month, Notion is free but doesn't cover time tracking."

Bad Example

"I want to build a task tracker" — too vague. AI will ask clarifying questions, but this wastes 5-7 messages (and 50-70 credits) to collect the same context. Specificity saves time and credits.

Step 3: Answer AI Questions

AI CPO conducts a structured interview, asking questions across five dimensions. Watch the context bar at the top of the chat — it shows progress for each dimension.

After each message, the platform automatically:

Extracts facts — LLM analyzes the message and identifies structured facts with weights (10-100)
Updates the context score — percentages for each dimension are recalculated
Extracts contacts — emails, phone numbers, Telegram handles are saved automatically
Unlocks artifacts — when enough data is gathered for the required dimensions
Suggests the next step — a hint appears in the context panel

Answer Strategy

Answer in detail — one long message is worth more than five short ones (and cheaper: 10 credits vs 50)
Quote real user words — "the client said: I lose an hour every day on tracking" (increases fact weight)
Name specific numbers — "3 out of 5 freelancers pay Toggl $10/month" (the fact gets a high weight)
Upload files — an interview transcript or chat export yields dozens of facts per message

Step 4: Generate Artifacts

When an artifact is unlocked (its icon in the sidebar becomes active), click on it. The preview panel shows information and a "Generate" button. Generation takes 10-30 seconds. The result renders with streaming (SSE).

Recommended Generation Order

Artifacts are interconnected — outputs from some are used as inputs for others. The optimal order:

Order	Artifact	Purpose	Required Context
1	Pain Map	Visualization of user problems	20%+
2	Job Statements	Job formulations in JTBD format	25%+
3	Segments (ABCDX)	Audience prioritization	35%+
4	Positioning	Competitive differentiation	40%+
5	Unit Economics	Financial model	50%+

Tip

You can generate artifacts in any order, but following the recommended sequence means each subsequent artifact will build on data from previous ones and be more accurate.

Step 5: Iterate

Artifacts are not static documents. They live and update along with your context:

Add information in chat — new facts from interviews, analytics data, competitor reviews
The system detects staleness — after each message, the MD5 hash of facts is checked; if it changed, the artifact is marked with a yellow "Data updated" banner
Regenerate with one click — costs 100 credits (half the price of initial generation)
Try a different model — the same artifact can be generated with Claude Sonnet, Gemini, or Qwen for comparison

Iteration Cycle

Conducted 3 interviews → uploaded transcripts → AI extracted 20 new facts → artifacts became stale → regenerated → now segments and positioning are based on real data, not hypotheses.

Step 6: Use the Feedback System

After generating each artifact, a feedback panel appears:

Accuracy — how correct are the claims in the artifact
Completeness — are all aspects covered
Relevance — how useful is this for your product

Click thumbs down on any dimension — a comment field appears. Feedback is auto-saved (1.2 second debounce) and used to improve prompts. If an artifact is inaccurate — write in the comment what exactly is wrong, add data to the chat, and regenerate.

Step 7: Export Results

Click the "Export" button in the top bar:

Pitch Presentation — investor slides compiled from key artifacts
PDF Report — full report with all artifacts (coming soon)
One-pager — one-page product summary (coming soon)

Quick Start Checklist

Step	Action	Result	Credits
1	Detailed product description	Context 15-25%	10
2	Answer 5-7 AI questions	Context 40-60%	50-70
3	Generate pain map + Job Statements	First 2 artifacts	400
4	Generate segments + positioning	Strategic artifacts	400
5	Launch auto research	Additional insights	500
6	Regenerate updated artifacts	Refined versions	200-400
Total			~1,500-1,800

Previous Platform Overview Next Research