Restaurant AI: menus, allergens, costing in one assistant

A restaurant's most painful institutional memory is the chef's. Recipes, allergen rules, costing decisions, vendor swaps — all in the chef's head. When the chef is on vacation, the kitchen runs slower; when the chef leaves, half the menu becomes a mystery.

A SeekFiles AI assistant scoped to your menu + recipe + allergen documents fixes this. Not by replacing the chef — by making their knowledge searchable.

What to upload

• Master recipe book (PDF or DOCX export from your POS / recipe software)
• Allergen matrix (which dishes contain which allergens)
• Vendor / ingredient catalog (with prices)
• Plating photos with notes
• Wine pairing notes
• Special dietary alternatives (vegan, gluten-free, halal, etc.)
• Past menu PDFs for reference

Build a "Kitchen Assistant" scoped to all of it. Optionally: "Front of House Assistant" scoped to allergen + service-facing docs only.

Real questions

Front of house

• "Is the carbonara safe for someone with a shellfish allergy?"
• "What gluten-free dessert options do we have today?"
• "What's the dairy in our chocolate cake?"

The answer comes with a citation to the allergen matrix, so the server can read the source and confirm with the kitchen if anything's unclear.

Kitchen

• "What's the full prep for the duck confit, including overnight cure?"
• "What's our standard plate weight for the burger?"
• "What's the sub if we're out of fresh basil?"
• "Cost out a 12-cover service of the tasting menu at current vendor prices."

Manager

• "Which dishes have the highest food cost percentage?"
• "Which menu items use the imported saffron — we need to swap vendors."
• "How many recipes use product X — is the per-unit cost increase worth absorbing or do we re-engineer?"

These are real questions managers ask once a quarter that take a half-day to answer manually. The Assistant answers in seconds.

Operational reality

• Update on menu change. New menu → upload new recipe book → next chat reflects.
• Allergen accuracy matters more than anything. Have the chef physically verify the allergen matrix before uploading. AI doesn't know what shellfish lives in your sauces.
• Don't depend on AI for the final allergen answer to a guest. Server reads it, walks to the kitchen, confirms. The Assistant is a fast first lookup; the kitchen is the source of truth.

Where it shines

• Pre-service brief. Quickly review tonight's specials, allergen flags, plate notes.
• New hire onboarding. "Ask the Assistant" beats "shadow a senior for 4 weeks" for routine questions.
• Costing scenarios. "If we swap vendor on tomatoes, what's the menu impact?" — much faster than re-spreadsheeting.

Where it fails

• Live operational decisions. A 6pm rush 86-out isn't an AI problem.
• Sensory judgment. AI can't tell you the stock tastes thin.
• Vendor relationships. Knowing who'll do you a favour on short notice — that stays in the manager's head.

A note on confidentiality

Recipes are intellectual property. Use the privacy-tier subscription that excludes training. Don't upload trade-secret recipes to a free-tier general chatbot.

Public Q&A bot for guests (optional)

A simplified Assistant scoped to allergen + menu basics, with the chef's approval, can be a public Q&A bot for guests to consult before booking ("can you accommodate a gluten + dairy + nut allergy?"). Reduces back-and-forth calls, improves accessibility.

The restaurants in our pilot dropped allergen-related front-desk calls by 60% with a public Q&A bot. The other 40% still call because they want a human voice — which is also fine.