Platform Overview
What is AI Product Builder?
AI Product Builder (AI CPO) is a platform for building and growing products based on the Product DNA (Discovery — Navigation — Acceleration) methodology. You chat with an AI assistant, describing your product, idea, or problem, and the platform automatically extracts facts, builds a Job Graph, generates 25 product artifacts, and helps you make data-driven decisions.
Unlike templates or document builders, AI CPO understands your product's context. Every artifact is built from collected facts about your niche, user pains, audience, economics, and competitors. The more context you provide, the more accurate the results.
How to Use the Platform
Three Workspace Panels
The workspace consists of three panels. You can resize them by dragging the dividers between them:
-
Sidebar (left) — artifact navigation grouped by phase:
Research, Strategy, Build, Launch, Diagnosis.
Also includes project switching, credit indicator, and the custom artifact request button.
Artifacts display colored status icons:
- 🔒 Gray — locked (insufficient data)
- 🔓 Blue — unlocked, ready for generation
- ✅ Green — generated
- ⚠️ Yellow — stale (data updated after generation)
- Preview (center) — view generated artifacts. You can open multiple artifacts in tabs, copy text, leave feedback (accuracy/completeness/relevance), and regenerate with a different AI model.
- Chat (right) — the primary interaction tool. Here you chat with the AI, upload files, launch research, and connect external data sources via connectors. At the top of the chat — the context bar and quick action buttons (research, connectors, backlog).
Getting Started
- Describe your idea or product — write in chat what you're building, for whom, and what problem you're solving
- AI extracts facts — each message is analyzed by an LLM that identifies facts across 5 dimensions
- Artifacts unlock — when enough data is collected, artifacts become available for generation
- Generate and iterate — click on an artifact, launch generation, review the result in the preview
- Add more context — the more information you provide, the better all artifacts become
Two Use Cases
UC1: New Idea — From Zero to MVP (Research Mode)
You have a product idea but no data. AI CPO will guide you through the entire journey: from niche research and problem formulation to landing page creation and pitch deck. At each step, the platform asks the right questions and generates artifacts that typically require a 3-5 person product team.
UC2: Existing Product — Diagnosis and Growth (Diagnosis Mode)
You already have a product but don't understand why users churn or how to grow. AI CPO will conduct a product audit, churn diagnosis using the Switch formula "in reverse" (why users "fire" your product), compare with industry benchmarks, and propose a growth strategy.
Context Score: 5 Dimensions
At the top of the chat, you'll see a context score — the percentage of collected context. Context is gathered across five dimensions. Each fact extracted from your message has a weight from 10 to 100. The sum of weights determines the context level.
| Dimension | What's Collected | Example Facts |
|---|---|---|
| Niche | Market, industry, geography, trends | "B2B SaaS for restaurants", "Market growing 20% per year" |
| Pains | Problems, frustrations, triggers | "Losing 5 hours per week on manual tracking", "Report errors" |
| Audience | Segments, roles, contexts, execution criteria | "Freelancers earning $1,000-3,000/month", "Criterion: speed" |
| Economics | Prices, WTP, budgets, business model | "Competitors charge $20/month", "WTP: $5-10/month" |
| Competitors | Alternative solutions, their strengths/weaknesses | "Toggl — strong in tracking, weak in reports", "40% don't use anything" |
Credit System
Every action on the platform costs a certain number of credits. When you create your first project, you receive 15,000 welcome credits — enough for a full analysis of several products.
| Action | Cost | Comment |
|---|---|---|
| Chat message | 10 credits | Each message = fact extraction |
| Artifact generation | 200 credits | First generation of any artifact |
| Artifact regeneration | 100 credits | Regeneration after data update |
| Premium artifact | 300 credits | Complex artifacts (unit economics, audit) |
| Research | 500 credits | Automated search via Telegram, web |
| Connector import | 300 credits | PostHog, GA4, Telegram, Yandex.Metrica |
| Custom artifact | 400 credits | Artifact by your request |
AI Models
The platform uses multiple AI models for different tasks:
| Task | Primary Model | Fallback |
|---|---|---|
| Chat (standard) | Qwen3-235B | GPT-OSS-120B → Nemotron |
| Chat (escalated) | Claude Sonnet 4 | Claude Haiku 4.5 |
| Fact extraction | Llama-3.3-70B (Groq) | Claude Haiku 4.5 |
| Artifact generation | Gemini 2.0 Flash | Qwen3-235B |
| Niche research | Llama-3.3-70B (Groq) | Gemini 2.0 Flash |
When generating artifacts, you can choose the model manually: Gemini 2.0 Flash (default), Qwen3-235B, Claude Sonnet 4, Claude Opus, or Claude Haiku. Model choice affects quality but not credit cost.
Feedback System
After generating each artifact, a feedback panel appears with three dimensions: Accuracy, Completeness, and Relevance. Give a thumbs up or down for each dimension. For negative ratings, a comment field appears.
Your feedback doesn't disappear — it's aggregated and analyzed by AI weekly. Based on the analysis, the platform automatically improves prompts and artifact generation. See Platform Evolution for details.
File Uploads
Drag a file into the chat area or click the paperclip button. Supported formats:
- Documents: PDF, DOCX, XLSX, CSV, TXT, JSON
- Images: PNG, JPG
- Audio: MP3, WAV, OGG, M4A, WEBM
Maximum file size is 10 MB. Files are validated by extension and magic bytes, renamed to UUID for security. File contents are analyzed by AI and facts are extracted automatically.
- Telegram chat exports from your target audience — AI will extract real pains from conversations
- Competitor reviews (App Store, Google Play) — for competitive analysis
- Customer development interview transcripts — maximum valuable facts
- Financial data (CSV) — for accurate unit economics
- Interview audio recordings — AI will transcribe and extract facts
Connectors
Connect external data sources for automated import and analysis. Click the API button in the chat toolbar.
| Source | Status | What's Imported |
|---|---|---|
| Telegram | Active | Posts from public channels |
| PostHog | Active | Trends, retention, funnels, DAU |
| Google Analytics 4 | Active | Traffic sources, demographics, conversions |
| Yandex.Metrica | Active | Visits, sources, behavior |
| VK, App Store, Google Play, etc. | Planned | Reviews, metrics |
Each import costs 300 credits. Data is depersonalized before analysis.
Escalation and Model Switching
If the AI detects dissatisfaction with responses (negative tone, repeated questions, no progress), it automatically switches to a more powerful model. The system detects 5 types of signals:
- Explicit request — "call a senior", "connect an expert"
- Negative tone — frustration, dissatisfaction in messages
- Stagnation — 4+ messages without new facts (context not growing)
- Repetition — more than 60% word overlap between messages
- Tone mismatch — serious question, superficial answer
Upon escalation, a notification appears in the chat: "Senior Expert connected" (empathetic, deep) or "Analyst connected" (data-driven, precise). When the conversation normalizes, AI automatically returns to the standard model to save resources.
Shadow Projects (No Registration)
You can start working without registration. The platform creates a "shadow" project tied to your session. All messages, facts, and artifacts are stored in the database (not in cookies). When you register or log in, the project automatically binds to your account — nothing is lost.