The Restaurant Owner Nobody Expected
I am not a chef.
I am a Chemical Engineer with four decades in the oil and gas industry. I have designed refinery systems, led multinational operations, and negotiated contracts across a dozen countries. I have spent my career in a world of pipelines, process simulations, and thermodynamic calculations.
So when I decided to open Level Twentynine — a Peruvian-Asian fusion restaurant and omakase sushi bar in Pembroke Pines, Florida — most people assumed I would hire a team of culinary consultants, bring in a brand agency, and spend months building everything the traditional way.
I did not.
Instead, I sat down with artificial intelligence and built the entire restaurant concept from the ground up.
The menu. The recipes. The food cost model. The market analysis. The business plan. The marketing strategy. The training materials. The brand artwork. All of it.
This is the story of how that happened — and how it changed everything about the way I think about building a restaurant.
Why AI? The Engineer’s Answer.
Engineers think in systems. A restaurant, at its core, is a system — inputs, processes, outputs, feedback loops. A menu is a product catalog with cost structures. A recipe is a technical specification. A food cost model is a financial engineering exercise.
When I looked at Level Twentynine through that lens, AI was not a shortcut. It was the right tool.
The question was never can AI help me build a restaurant? The question was how do I ask the right questions?
That skill — the ability to frame a precise prompt, challenge an output, refine, iterate, and arrive at something excellent — is not so different from engineering. And it turned out to be the most important skill I brought to this project.
The AI Journey: From ChatGPT to Claude
My AI-assisted restaurant development did not begin with a single tool or a single day. It evolved.
The first menus I developed using artificial intelligence were created in early 2025, using ChatGPT. That process taught me how to think in prompts — how to translate culinary vision into language an AI could work with, and how to evaluate and refine the outputs I received.
But as the project deepened and the demands became more sophisticated, I moved to Claude by Anthropic. The menus currently in use at Level Twentynine in 2026 — the Nikkei kitchen menu, the Kaito Room omakase menu, and the Torii Brunch concept now in development — were all built using Claude.
The difference was not just capability. It was the quality of the dialogue. Claude understood context, held complex instructions across long conversations, and produced outputs that felt genuinely crafted rather than assembled. For a project with the depth and nuance of a Nikkei restaurant concept, that distinction mattered enormously.
The Data-Driven Menu: Starting with What Guests Were Already Telling Me
Before writing a single new menu item, I did something most restaurant owners skip entirely: I went to the data.
I uploaded Level Twentynine’s Toast sales summary from the prior year into Claude. What came back was not just a spreadsheet analysis — it was a comprehensive understanding of the restaurant’s actual performance. Total gross sales. Discounts applied. Number of guests served. And most importantly: the top 50 most popular dish items ordered by real guests over a full year of operations.
This changed everything about how I approached the new menu.
Instead of guessing what guests loved, I knew. I could see precisely which dishes had earned their permanent place on the menu and which were quietly underperforming. The data gave me the confidence to make two decisions that no amount of culinary intuition alone could have produced with the same certainty: which dishes to keep, and which to retire.
From that foundation, I layered in a second dimension — culinary integrity. Alongside the top sellers, I identified the iconic Peruvian, Nikkei, and Chifa dishes that any serious restaurant of this tradition must offer, regardless of their sales volume. A menu without a proper Lomo Saltado or a classic Corvina Ceviche is not a Peruvian menu. Some dishes earn their place not through sales data but through cultural obligation.
The result was a menu built on two pillars: what guests had proven they wanted, and what the cuisine demanded in terms of authenticity and identity.
Building the Menu: Depth, Architecture, and a Real Conversation
The Level Twentynine menu covers Nikkei, Chifa, and Sushi across appetizers, ceviches, tiraditos, hot dishes, rolls, omakase tracks, desserts, and cocktails. With the data analysis complete and the culinary framework defined, I worked with Claude to architect the full menu structure.
This was not a single prompt producing a final output. It was a sustained dialogue. I pushed back on dishes that felt too generic. I asked for more authentic Nikkei preparations. I challenged the balance between accessibility and sophistication — this is Pembroke Pines, not Miami’s South Beach, and the pricing and positioning needed to reflect that reality. I requested that every dish carry a description worthy of the story behind it.
The result was a menu built with internal logic: price architecture calibrated to the local market, flavor profiles that built on each other across courses, and a dual omakase track in the Kaito Room — the Kaito Terroir featuring Florida and Gulf seafood, and the Kaito Prestige showcasing Japanese imports.
Additionally, the Torii Brunch — Level Twentynine’s weekend brunch concept, currently a work in progress — is being shaped through the same AI-assisted process. While still in development, its structure, menu architecture, and brand identity are already taking form through Claude. When it launches, it will carry the same depth of thought and cultural intentionality as everything else in the portfolio.
The Recipes: Precision Meets Tradition
Once the menu existed, I needed something more: actual recipes. Standardized. Scalable. Teachable to a kitchen team that would need to produce consistent results night after night, service after service.
This is where AI proved genuinely extraordinary.
I worked with Claude to develop full recipe specifications for each dish — ingredients, quantities, preparation method, plating notes, and the cultural context behind each preparation. For the Nikkei dishes, I requested historically grounded recipes that respected the Japanese-Peruvian tradition. For the Chifa dishes, I wanted the authenticity of Chinese-Peruvian cooking rooted in the calle capón kitchens of Lima.
What I received were not generic internet recipes. They were thoughtful, layered preparations that reflected real culinary knowledge — knowledge I then reviewed with my chefs, refined based on their professional input, and codified into the kitchen’s operating standards.
The AI gave me a first draft that any culinary team could build upon. That is not a small thing. That is months of recipe development compressed into days.
The Food Cost Model: Three Months Versus Twenty Minutes
I want to be direct about this one, because the contrast is striking enough that it deserves its own moment of honesty.
Before using AI, I built Level Twentynine’s original food cost spreadsheet myself. Manually. It took me approximately three months — three months of inputting ingredients, calculating portion costs, building formulas, checking margins, and maintaining a living document that reflected the realities of the kitchen.
I am not complaining about those three months. The process taught me the economics of the restaurant in a way that no consultant’s report ever could have. But three months is three months.
When it came time to build the new, expanded food cost model for the 2026 menu, I uploaded the menu PDFs directly into Claude and asked it to build a complete cost workbook. What I received — in minutes — was a nine-sheet Excel model covering all 108 menu items, with ingredient costs, portion sizing, a 4x cost multiplier for menu pricing, color-coded food cost percentage thresholds by category, and a summary dashboard.
Three months. Then minutes.
The model gave my operations team a financial engineering tool — not a simple spreadsheet. It allowed us to see at a glance which dishes were performing within healthy margins and which needed repricing or recipe reformulation. It created the kind of operational clarity that makes the difference between a restaurant that survives its first three years and one that thrives.
Beyond the Menu: A Full Business Infrastructure Built with AI
What began as a menu development project became something much larger.
Using the sales data, the menu analysis, and the culinary framework already developed, I worked with Claude to build out the complete operational and strategic infrastructure for Level Twentynine. This included:
A market analysis examining the Pembroke Pines and broader South Florida competitive landscape, identifying positioning opportunities and pricing benchmarks against comparable Nikkei, fusion, and omakase concepts in the region.
A competitive price analysis ensuring that Level Twentynine’s menu pricing reflected both the quality of the offering and the economic realities of its local market.
A demographic study of the restaurant’s trade area, informing decisions about menu composition, language, service style, and marketing channel priorities.
A full business plan and marketing plan that articulated the restaurant’s value proposition, revenue model, target customer, and growth strategy.
Training materials for both Back of House and Front of House personnel — standardized, detailed, and aligned to the menu and service philosophy of the restaurant.
Every one of these documents was produced through a structured AI-assisted process. Not as a replacement for judgment, but as an accelerant of it.
The Brand Art: Scripting Midjourney, Plating Through Photography
A menu without visual identity is a specification sheet. Level Twentynine needed to feel like something.
Here is where the process became genuinely creative — and where I discovered something unexpected about the relationship between language and image.
I used Claude to write the prompts — the detailed, evocative scripts — that I then fed into Midjourney, the AI image generation platform, to produce visual content. Claude did not generate the images. Claude wrote the language that told Midjourney exactly what to create.
This matters more than it might seem. A Midjourney prompt is not a casual description. It is a directed creative brief — specifying lighting, mood, composition, cultural aesthetic, color temperature, and emotional tone simultaneously. Writing a great prompt requires the same precision as writing a great recipe. Claude helped me develop that precision systematically, dish by dish.
The results served two purposes simultaneously. For the menu and marketing materials, they gave Level Twentynine professional-grade food photography at a fraction of traditional production costs. For the kitchen, they served as plating guides — visual references showing the Back of House team exactly how each dish should look when it leaves the pass.
A photograph is worth a thousand words of plating instruction. AI made those photographs possible.
But the image that stopped everyone who saw it — the one that captures the entire soul of Level Twentynine in a single frame — was not a food photograph.
It was a landscape.
Where Machu Picchu Meets Mount Fuji and the Great Wall of China.
Three civilizations. Three culinary traditions. Three world wonders, united in a single sweeping image that tells the story of Nikkei cuisine without a single word. Peru. Japan. China. The three cultures whose flavors, histories, and peoples converged to create the cuisine that defines this restaurant.
That image now represents something beyond a menu. It is the brand’s creation myth rendered visible.
What I Learned: Seven Principles of AI-Assisted Restaurant Development
After building Level Twentynine’s entire concept infrastructure with artificial intelligence, here is what I know:
1. AI does not replace vision. It accelerates it.
The concept — Nikkei, the Kaito Room, the Pembroke Pines market positioning — came from me. AI gave that vision structure, depth, and speed that no traditional process could have matched.
2. Data is the foundation. AI is the architect.
Loading real sales data from Toast before building the new menu changed everything. The AI did not guess what guests wanted — it analyzed what they had already proven they wanted. That distinction is the difference between a menu built on instinct and one built on evidence.
3. The quality of your output is the quality of your input.
A vague prompt produces a vague result. A precise, culturally informed, strategically framed prompt produces something you can actually use. Learning to prompt well is a genuine professional skill.
4. AI is a collaborator, not an oracle.
Every output I received, I challenged. I pushed back, refined, and iterated. The final products were the result of a real dialogue between my knowledge and the AI’s capability.
5. Domain expertise multiplies AI’s value.
Because I understood food cost structures, process optimization, and brand strategy, I knew what to ask for and how to evaluate what I received. AI is most powerful in the hands of someone who knows their field.
6. AI compresses time without compressing quality.
Three months of manual spreadsheet work. Minutes with Claude. That gap does not mean the work was less rigorous. It means the time was redirected — from data entry to decision-making.
7. The human standard is still the final standard.
AI built the scaffold. My chefs, my team, and my guests determine whether what stands on that scaffold is worth returning to. Technology can build a menu. Only hospitality can build a restaurant.
The Restaurant Is Real. The Method Is Replicable.
Level Twentynine is open. The Kaito Room is serving omakase. The food cost model is running. The brand exists in the world — and somewhere in the dining room, a guest is looking at a menu photograph of a dish they are about to order, not knowing that the image was written by one AI and painted by another.
The Torii Brunch is coming. Another chapter, built the same way.
If this story tells you anything, I hope it tells you this:
The tools available to entrepreneurs today are extraordinary. The discipline to use them well is the only competitive advantage that cannot be automated.