Agentic AI Strategy Analyzer

What It Does

The Agentic AI Strategy Analyzer takes any work or business scenario and produces a structured analysis of where AI agents could create value. Describe a situation — staffing operations, customer onboarding, research workflows — and the tool breaks it down into specific agent opportunities, system architectures, and implementation paths.

This isn't a chatbot conversation. It's a structured analytical tool that produces a complete strategy document in real time.

How It Works

Streaming Analysis

The analyzer sends your scenario to a large language model via OpenRouter, configured with a specialized system prompt that enforces structured analytical output. Responses stream back in real time using Server-Sent Events — you see the analysis forming as the model reasons through your scenario.

Structured Output Sections

Every analysis follows a consistent structure:

• Understanding Your Business — The model reflects back its interpretation of your scenario, identifying the core operational challenges and scale factors
• Top 3 AI Agent Opportunities — Specific, named agents with descriptions of what they do, expected impact, and implementation complexity
• Agent Architecture — A mermaid flowchart showing how the proposed agents connect, what data flows between them, and where they integrate with existing systems
• Opportunity Map — A knowledge tree visualizing the full landscape of AI opportunities across your scenario, organized by domain
• Implementation Roadmap — A phased plan from quick wins to full system deployment

Architecture Diagrams

The analyzer generates two types of visual diagrams using Mermaid.js:

Agent Architecture Diagrams are top-down flowcharts (graph TD) showing the multi-agent system design — orchestrator agents, specialized workers, data sources, and output systems connected with directional flows.

Opportunity Maps are left-to-right knowledge trees (graph LR) that branch from the core scenario into domains, each splitting further into specific agent opportunities.

Both render as SVGs with a dark theme matching the site's design language.

Technical Architecture

Frontend: React client component handles the streaming connection, accumulates response chunks, and renders via react-markdown with a custom component pipeline that detects mermaid code blocks and routes them to the MermaidDiagram renderer.

API Layer: Next.js API route receives the scenario payload, constructs the system prompt, and POSTs to OpenRouter's chat completions endpoint with streaming enabled. The response is piped back as Server-Sent Events.

Diagram Rendering: Mermaid.js is dynamically imported (it's a large library) and initialized with custom theme variables matching the site's design tokens. Each mermaid code block is rendered to SVG programmatically using mermaid.render().

Session Logging: Each analysis session is logged to Supabase with the full conversation, context metadata, and optional lead email for users who want to stay updated.

Why We Built It

The Analyzer serves two purposes. First, it demonstrates the kind of structured analytical reasoning that agentic AI systems can produce — moving beyond conversational Q&A into real strategic output with visual artifacts. Second, it's a useful tool on its own. Anyone exploring how AI agents could apply to their operations gets a concrete, specific starting point.

It's also a testing ground for techniques we're exploring in our research: structured prompting, streaming architectures, and rendering complex AI output as interactive visual content.