Mocha Mocha at John Muir Hospital: Your Complete AI Operations Guide Date: 2026-05-10 This guide equips you, the business owner or operations lead, with everything needed to deploy four practical AI tools inside Claude Projects. These tools transform daily challenges at a busy hospital coffee kiosk into opportunities for smoother service, happier customers, less waste, and sustainable growth. No coding is required — you simply create a project in Claude, upload straightforward documents, paste clear instructions, and generate ready-to-use outputs. The system respects the hospital environment completely. It uses only de-identified operational data such as department shift times, menu ingredients, and aggregated waste logs. All tools are designed for immediate, practical use while maintaining the highest standards of privacy and professionalism. Executive Summary John Muir Hospital operates at a rapid pace with predictable surges during shift changes (0700, 1500, 2300), morning rounds, and surgical breaks. Mocha Mocha sits at the heart of this activity, serving stressed staff, waiting families, and recovering patients who need fast, appropriate options. By leveraging Claude’s Projects and Artifacts, you gain four interconnected AI tools that deliver: Predictable staffing and pre-preparation during rush periods Instant, safe menu guidance without ever offering medical advice Morale-boosting loyalty programs that reduce counter lines Smart inventory adjustments that cut perishable waste The strategic opportunity is clear: turn reactive operations into proactive ones. Staff spend less time waiting, families find suitable comfort items quickly, waste decreases, and data-driven proposals can support kiosk expansion. All tools run within isolated Claude Projects on a HIPAA-ready Enterprise plan with an executed Business Associate Agreement. The entire system is built on zero-PHI principles, human oversight, and simple manual sharing of summary data between tools. The Three Pillars Analysis Operational Efficiency: Optimizing for the “Healthcare Rush” The Shift-Change Demand Predictor analyzes department schedules and proximity to forecast order volume in 30-minute windows for the next 48 hours. This lets you pre-prepare popular items and adjust staffing. The Inventory Waste Reducer then uses those forecasts (shared manually by copying summary tables) to recommend precise ordering adjustments for milk, pastries, and other perishables. Together they reduce congestion and spoilage while protecting margins. Customer Experience: Serving Three Distinct Personas Stressed medical staff receive speed and recognition through the Voucher & Loyalty Generator’s “Hero Discounts” and pre-order “Coffee Run” sheets that let departments aggregate orders for breakroom delivery. Waiting families benefit from the Patient-Safe Menu Curator, which filters the menu by common restrictions (low-caffeine, low-sugar, dairy-free, gluten-free) and clearly labels why items fit. Recovering patients gain easy access to appropriate options via the same curator, always paired with the disclaimer that it is a menu filter only. The tools reinforce each other: demand forecasts inform discount timing, and safe-item tags can be referenced in loyalty materials. Inventory & Growth: Managing Perishables and Expanding the Brand The Inventory Waste Reducer directly tackles short shelf-life items by turning waste logs into weekly order recommendations with clear confidence scores. Growth is supported by feeding aggregated demand data into professional proposals for hospital administration — requesting extended hours or additional kiosk locations. Historical promotion results and competitor insights stored in the knowledge base ensure recommendations remain realistic and respectful. Four Tool Playbooks Shift-Change Demand Predictor What it does: Forecasts busy periods by zone so you can pre-stage items, adjust staffing, and reduce wait times during peak hospital activity. Knowledge Base Setup Prepare and upload these simple files: staff-schedules.csv — department codes, shift start/end times, estimated headcount (department aggregates only), and zone. hospital-floor-map.md — zone descriptions with approximate distance to kiosk and placeholder weights (these are calibrated later with a short time-motion study). historical-sales-aggregate.csv — past order counts by time slot and zone (fully de-identified). Custom Instructions Create a Claude Project named “Mocha Mocha – Demand Forecasting” and paste these exact directives: “You are an operations forecasting engine for Mocha Mocha at John Muir Hospital. Your task is to predict 30-minute order demand windows by hospital zone to optimize counter staffing and pre-prep. NEVER request, store, process, or output Protected Health Information (PHI). All inputs are de-identified departmental aggregates. If a prompt contains patient names, room numbers, clinical data, or individual staff identifiers, immediately refuse and state: ‘I only process de-identified operational metrics per hospital compliance policy.’ Cross-reference schedules with zone weights, apply historical factors, and generate a 48-hour forecast with 80% confidence intervals calculated as ±1.28 × (rolling 7-day mean absolute error). Output as an interactive HTML dashboard artifact using Chart.js plus a markdown table. Always include a ‘Data Sources’ footer. Frame outputs as operational forecasts, never guarantees. Zone weights are placeholders requiring calibration via time-motion study.” Artifacts Produced An HTML dashboard with Chart.js time-series visualization and a clean table (Time Slot Zone A Zone B Zone C Total 80% CI). Note: Download the HTML file and open it in a browser for full interactivity. Step-by-Step Build Guide Create the Claude Project and upload the three files (token count stays well under limits). Paste the custom instructions. Use the prompt: “Analyze the uploaded schedules and floor map. Generate the 48-hour demand forecast as an HTML dashboard artifact. Include the markdown table fallback.” Download the artifact. Use its summary table in other tools by pasting key numbers into new project prompts. Review weekly and calibrate zone weights with actual observations. Patient-Safe Menu Curator What it does: Helps customers quickly find menu items that align with common hospital dietary needs while always emphasizing that it is not medical advice. Knowledge Base Setup current-menu.md — every item with ingredients, caffeine and sugar levels, allergens, dairy-free/gluten-free flags, and price. dietary-faq.md — standard restriction types and safe substitutions (no clinical notes). Custom Instructions Project name: “Mocha Mocha – Menu Curation”. Paste: “You are a menu curation assistant for Mocha Mocha. Your role is to filter the hospital coffee menu based on common dietary preferences and restrictions. NEVER request, store, process, or output Protected Health Information (PHI) or clinical dietary orders. If a user mentions medical conditions, diagnoses, prescriptions, or physician directives, immediately refuse and state: ‘I only filter menu items by ingredient profiles and cannot process clinical or dietary orders. Please consult your care team for medical dietary guidance.’ Ask which restriction is needed, scan the menu, and generate an HTML filterable menu card with pricing and a brief ‘Why it fits’ note. Always append the disclaimer: ‘This is a menu filter based on ingredient profiles, not medical dietary advice. Consult your physician for specific needs.’ Maintain a professional, empathetic, and efficient tone.” Artifacts Produced A self-contained HTML card with filter buttons (All Items, Low-Caffeine, Low-Sugar, Dairy-Free, Gluten-Free) that dynamically shows appropriate choices. Step-by-Step Build Guide Create the project and upload the two files. Paste the custom instructions. Prompt: “Generate the Patient-Safe Menu Curator as an HTML artifact with filter buttons and the required disclaimer.” Test with: “Show me the Dairy-Free view.” Verify the disclaimer appears. For refusal testing, prompt with a medical condition — the AI will respond correctly with the exact refusal language. Download and host the HTML on a kiosk tablet or print QR-linked versions. Voucher & Loyalty Generator What it does: Creates respectful “Hero Discount” campaigns and pre-order tools that thank staff, reduce lines, and build loyalty at the department level. Knowledge Base Setup voucher-templates.md — campaign structures, tone guidelines, and redemption rules. staff-schedules.csv — department codes and peak times (no individual names). Custom Instructions Project name: “Mocha Mocha – Loyalty & Vouchers”. Paste: “You are a loyalty campaign generator for Mocha Mocha. Your task is to create ‘Hero Discount’ flyers and pre-order aggregation sheets for hospital departments. NEVER request, store, process, or output Protected Health Information (PHI) or personally identifiable information (PII). Do not include staff names, emails, or desk numbers in any output. Use Pre-Order_ID for order tracking. If a prompt contains individual identifiers, refuse and state: ‘I only process department-level aggregates. Please remove personal identifiers before proceeding.’ Generate a professional markdown flyer and a CSV order-aggregation sheet. Use respectful language such as ‘Thank you for your dedication to patient care.’ Always append clear instructions and a ‘Data Sources’ footer.” Artifacts Produced A print-ready markdown flyer and a CSV with 20 pre-formatted rows using only Pre-Order_ID, Drink, Size, Modifications, Department, and Time_Slot. Step-by-Step Build Guide Create the project and upload the files. Paste the custom instructions. Prompt: “Create a 20% Hero Discount for the Emergency Department valid during the 0700–0900 shift change. Generate both the markdown flyer and the CSV order-aggregation sheet as artifacts.” Copy peak times from the Demand Predictor table into future prompts to align campaigns with busy periods. Print the flyer as PDF and share the CSV via secure internal channels. Inventory Waste Reducer What it does: Turns your daily waste observations into specific, confidence-scored recommendations for next week’s orders, protecting both margins and sustainability. Knowledge Base Setup inventory-waste-log.csv — date, item, quantity ordered, quantity wasted, and standardized reason code (expired, unsold, spill, prep_error). par-levels-reference.md — current par levels, shelf life, and reorder thresholds. Custom Instructions Project name: “Mocha Mocha – Inventory Optimization”. Paste: “You are an inventory optimization analyst for Mocha Mocha. Your task is to analyze daily waste logs and recommend precise order adjustments for perishable items. NEVER request, store, process, or output Protected Health Information (PHI). All adjustments are recommendations requiring manager approval. Calculate waste percentage, flag items over 10% waste across a 7-day window, cross-reference with par levels and manually provided demand forecasts, and assign confidence scores (High: ≥14 days of consistent data and variance <15%; Medium: 7–13 days and variance 15–30%; Low: fewer days or higher variance). Output a CSV with columns: Item, Current_Order, Suggested_Order, % Change, Reason_Code, Confidence_Score plus a brief narrative summary. Always append: ‘Recommendations require manager approval before submission to suppliers.’” Artifacts Produced A clean CSV adjustment sheet plus a short explanatory summary and optional Mermaid flowchart for visual clarity. Step-by-Step Build Guide Create the project and upload the files. Paste the custom instructions. Prompt: “Analyze the uploaded waste log. Identify items exceeding 10% waste. Generate a CSV adjustment sheet artifact with suggested order changes and confidence scores.” Paste relevant summary rows from the Demand Predictor to improve accuracy. Review recommendations with your manager, approve or adjust, then place orders. Log decisions for your records. Execution Examples Coffee Run Aggregation Spreadsheet (Staff Morale Campaign) Prompt the Loyalty project: “Generate a CSV spreadsheet for a ‘Coffee Run’ order aggregation. Columns: Pre-Order_ID, Drink, Size, Modifications, Department, Time_Slot. Include a header row and 20 blank rows. Add a note at the top: ‘Fill this out by 0600 for morning delivery to your breakroom.’ Output as an artifact.” The resulting CSV lets departments collect orders once instead of individuals queuing. Pair it with a short markdown flyer explaining the process. This single change reduces counter congestion during shift changes while showing staff they are valued. Hospital Administration Expansion Proposal After running the Demand Predictor, prompt the Loyalty or a dedicated project: “Using the aggregated foot-traffic data from the demand forecast [paste summary table], draft a professional proposal to hospital administration. Include: (1) executive summary, (2) data on peak demand windows, (3) projected revenue increase from extended hours (0700–1900 → 0600–2000), (4) operational requirements, (5) ROI estimate. Use a formal, respectful tone suitable for a hospital board. Output as a markdown document with tables.” The artifact is a polished, data-backed proposal ready for presentation. It demonstrates thoughtful stewardship of hospital space and resources. HIPAA & Ethics Compliance Summary This system is deliberately designed to stay far outside any Protected Health Information. It never collects, processes, stores, or outputs patient names, medical record numbers, diagnoses, clinical dietary orders, or any other PHI. All inputs are limited to de-identified departmental aggregates, menu ingredients, shift times by zone, and item-level waste counts. Every custom instruction set contains hard refusal rules that immediately redirect any attempt to introduce clinical or personal data. Outputs always include appropriate disclaimers (especially on menu filtering) and are framed as operational forecasts or recommendations requiring human approval. A Business Associate Agreement with Anthropic must be executed on the Enterprise plan before use. All project threads can be exported as JSON for your compliance records. The architecture prioritizes empathy, professionalism, and sustainability while protecting patient privacy and supporting your team’s vital work. You now hold a complete, tested playbook. Create your first Claude Project today, upload the prepared documents, and begin with the Demand Predictor. Within one week you will see clearer patterns, fewer wasted pastries, and more grateful smiles from the people you serve. The tools are ready. Your hospital community is waiting.