AI-Powered Menu Intelligence & Proposal Generation
This digital worker orchestrates 9 specialized AI agents in a DAG (Directed Acyclic Graph) workflow to analyze uploaded menus using computer vision and NLP, gather competitive intelligence, match products to supplier catalogs using semantic embeddings, forecast demand patterns, assess sustainability impact, and auto-generate professional sales proposals. The workflow completes in 45-90 seconds with real-time progress visualization and full transparency into agent reasoning.
Problem Statement
The challenge addressed
Solution Architecture
AI orchestration approach
Workflow Configuration Screen - Restaurant profile setup with menu source selection, system health monitoring showing all services operational, and agent fleet displaying 9 AI agents ready for menu analysis workflow.
Live Multi-Agent Execution - Real-time workflow showing agents processing menu data at 25% completion, with Sustainability Analyzer actively running and live discoveries panel tracking 45 dishes, 116 ingredients, and 68 products identified.
Analysis Complete Dashboard - Results summary showing successful menu analysis for The Classic American Grill with 116 ingredients identified, 108 products matched (96% coverage), 88% win probability, and $1,101 monthly savings opportunity.
Performance Metrics & Observability - Technical dashboard displaying token usage, latency statistics, LLM cost breakdown by model, agent performance with confidence scores, and product matching results showing 87 exact matches and 15 fuzzy matches.
AI Agents
Specialized autonomous agents working in coordination
Workflow Coordinator
Coordinating multiple AI agents with complex dependencies and ensuring proper sequencing of tasks while maintaining workflow integrity and error handling.
Core Logic
Acts as the central coordinator (model: claude-3-5-sonnet-20241022, icon: hub, color: #6366f1) that manages the entire menu analysis workflow, routes tasks to specialized agents based on the DAG structure, synthesizes intermediate results, handles errors gracefully, and ensures all agents complete their work before finalizing the proposal. Utilizes tools: coordinate_workflow, route_tasks, synthesize_results, handle_errors.
NLP Specialist
Extracting structured data from unstructured menu images and documents with varying formats, fonts, and layouts.
Core Logic
Uses GPT-4 Vision (model: gpt-4-vision, icon: restaurant_menu, color: #10b981) to analyze menu images, identify dishes, extract prices, categorize items (appetizers, entrees, desserts), detect dietary labels, and output structured JSON with item names, descriptions, prices, and attributes. Handles handwritten menus, PDFs, photos, and scanned documents. Utilizes tools: analyze_menu_image, extract_text_structure, categorize_items, identify_dietary_labels.
Competitive Intelligence
Lack of competitive market insights when pricing and positioning menu items, leading to suboptimal margins and missed opportunities.
Core Logic
Analyzes market data, competitor pricing, regional trends, and seasonal patterns (model: claude-3-5-sonnet-20241022, icon: trending_up, color: #f59e0b) to provide competitive positioning insights. Identifies price optimization opportunities, market gaps, and trending items that can improve profitability. Utilizes tools: fetch_competitor_prices, analyze_regional_trends, identify_market_gaps, calculate_price_elasticity.
Real-Time Intelligence
Menu planning without awareness of supply chain disruptions, lead times, and ingredient availability leads to out-of-stocks and menu item unavailability.
Core Logic
Monitors real-time supply chain status (model: claude-3-5-sonnet-20241022, icon: local_shipping, color: #ef4444), tracks supplier lead times, identifies potential disruptions (weather, logistics, shortages), and provides availability confidence scores for menu items. Integrates with supplier APIs and logistics data feeds. Utilizes tools: monitor_supply_status, check_ingredient_availability, track_lead_times, assess_disruption_risk.
Predictive Analytics
Inaccurate demand predictions leading to over-ordering (waste) or under-ordering (stockouts) of menu item ingredients.
Core Logic
Applies ML-powered demand forecasting (model: claude-3-5-sonnet-20241022, icon: auto_graph, color: #8b5cf6) using historical sales data, seasonality patterns, event calendars, weather forecasts, and market trends. Provides item-level demand predictions with confidence intervals and optimal order quantity recommendations. Utilizes tools: predict_demand, analyze_seasonality, incorporate_events, calculate_confidence_intervals.
ESG Intelligence
Inability to quantify and improve the environmental sustainability of menu offerings and sourcing decisions.
Core Logic
Calculates carbon footprint for menu items (model: claude-3-5-sonnet-20241022, icon: eco, color: #22c55e) based on ingredients, sourcing distance, production methods, and packaging. Provides sustainability scores, identifies eco-friendly alternatives, and tracks ESG metrics for compliance and reporting. Utilizes tools: calculate_carbon_footprint, score_sustainability, identify_green_alternatives, generate_esg_report.
Semantic Search
Manually matching menu ingredients to supplier catalogs with thousands of SKUs is time-consuming and error-prone.
Core Logic
Uses vector embeddings (model: text-embedding-3-large, icon: search, color: #ec4899) to semantically match menu items to supplier product catalogs. Handles variations in naming, packaging sizes, and brands. Returns ranked matches with similarity scores, pricing, and availability from multiple suppliers. Utilizes tools: generate_embeddings, search_catalog, rank_matches, compare_suppliers.
Document Generation
Creating professional, data-driven sales proposals is time-consuming and often lacks the analytical depth needed to win deals.
Core Logic
Synthesizes insights from all agents (model: claude-3-5-sonnet-20241022, icon: description, color: #06b6d4) to generate comprehensive sales proposals including executive summary, cost analysis, product recommendations, savings projections, sustainability impact, and implementation timeline. Outputs formatted documents ready for client presentation. Utilizes tools: generate_proposal, format_document, create_visualizations, calculate_roi.
Quality Assurance
Ensuring accuracy and consistency of AI-generated analysis and proposals before delivery to stakeholders.
Core Logic
Performs validation checks on all agent outputs (model: claude-3-5-sonnet-20241022, icon: verified, color: #78716c), verifies data accuracy, checks for inconsistencies, validates pricing calculations, ensures proposal completeness, and flags any issues requiring human review. Utilizes tools: validate_data, check_calculations, verify_consistency, flag_review_items.
Worker Overview
Technical specifications, architecture, and interface preview
System Overview
Technical documentation
Tech Stack
6 technologies
Architecture Diagram
System flow visualization