Menu Engineering & Recipe Costing with AI

Your menu is the single most powerful profit lever in your restaurant. A 1% improvement in menu mix can be worth more than a successful Yelp campaign or a new shift of advertising. And yet most independent operators set prices once when they open and only revisit them when a supplier crisis forces their hand.

AI changes this. You can cost a recipe, set a price, write the menu copy, and analyze your top sellers in a single chat session. This lesson shows you how.

What You'll Learn

• How to use AI to cost a recipe and recommend a price
• How to write evocative, on-brand menu descriptions that sell
• How to do classic menu engineering (Stars, Plowhorses, Puzzles, Dogs) with AI
• How to write seasonal menu rewrites and LTO (limited-time offer) copy

Step 1: Cost a Recipe in 60 Seconds

A standard recipe-costing exercise — yields, conversions, applied prices — takes 15–30 minutes by hand. With AI, you describe the recipe and let it do the math.

Sample prompt:

Act as my food cost analyst.

Recipe: Bolognese sauce (yields 4 quarts, ~32 portions at 4 oz)

Ingredients (case prices):
- Ground beef 80/20 — $4.80/lb, 4 lb
- Italian sausage — $5.20/lb, 1 lb
- Yellow onion — $0.90/lb, 1 lb
- Carrot — $1.20/lb, 0.5 lb
- Celery — $1.10/lb, 0.5 lb
- San Marzano tomatoes — $4.50/28oz can, 2 cans
- Dry red wine — $11/750ml bottle, 1.5 cups
- Heavy cream — $4.20/qt, 1 cup
- Olive oil, herbs, salt — call it $3 total

Calculate:
1. Total batch cost
2. Cost per 4 oz portion
3. If I want a 28% food cost, what menu price?
4. If I want a 25% food cost, what menu price?

Show your math.

You'll get a clean cost-per-portion plus pricing options in seconds. Always sanity-check the math — but it will be right roughly 95% of the time on simple recipes.

Step 2: Write Menu Copy That Sells

Restaurant menu language has a published research base: descriptive menu items sell up to 27% more than identically priced plain ones. AI is very good at descriptive menu copy if you keep it from sliding into clichés.

Anti-cliché prompt:

[paste house context]

Write a menu description for our Bolognese:
- Beef and sausage
- Soffritto base
- San Marzano tomatoes
- 4-hour simmer
- Tossed with hand-rolled pappardelle, finished with
  pecorino and chili flake

Rules:
- Under 25 words
- No words: artisanal, elevated, curated, hand-crafted,
  delectable, mouthwatering, foodie, perfectly
- Use specific verbs and named ingredients
- Sound like a confident neighborhood Italian, not a
  hotel restaurant

Specific. Verb-driven. On-brand. That's the formula.

Step 3: Menu Engineering — Stars, Plowhorses, Puzzles, Dogs

Classic menu engineering plots every menu item on two axes: profitability (high/low) and popularity (high/low). The four quadrants:

• Stars — high profit, high popularity → keep, feature visually, never discount
• Plowhorses — low profit, high popularity → re-engineer cost, slight price bump, or upsell with sides
• Puzzles — high profit, low popularity → reposition, rename, or move to a better menu spot
• Dogs — low profit, low popularity → cut

Paste a 30-day item-level sales export from your POS into ChatGPT or Claude with a prompt like:

Act as my menu engineering consultant.

Below is 30 days of item-level sales data for my
40-seat Italian restaurant. Columns: item, count sold,
selling price, food cost.

[paste data]

For each item:
- Calculate gross profit per unit
- Calculate total monthly contribution
- Classify as Star / Plowhorse / Puzzle / Dog
  (use the 50th percentile in each axis as the cutoff)
- Give me one specific action recommendation

End with a summary table sorted by category.

You'll have an actual menu strategy in 90 seconds. A menu engineering consultant would charge $1,500–$3,000 for the same exercise.

Step 4: Seasonal Rewrites and LTOs

Every quarter you should refresh menu copy and consider seasonal LTOs. AI accelerates this dramatically.

Try:

[paste house context]

Generate 8 fall LTO ideas for our menu:
- 3 pasta dishes (squash, mushrooms, brassicas)
- 3 mains (game, root vegetables, slow braises)
- 2 desserts (apple, pear, fig, chestnut)

For each: dish name, one-sentence description,
estimated food cost % at $X price point, and
3-word Instagram pitch.

You'll get a quarterly menu meeting starter pack in two minutes.

Step 5: Allergen and Dietary Tagging

Most modern POS and online ordering systems require allergen tagging. AI does this in bulk:

For each menu item below, tag for:
GF (gluten-free), V (vegetarian), VG (vegan),
DF (dairy-free), N (contains nuts), GFA (gluten-free
adaptable).

[paste menu]

Output as a markdown table.

Always have your chef double-check before publishing — but the first pass is done.

Real-World Time Savings

A single Monday morning of menu work that used to take 4 hours:

• 6 new dishes costed → 8 minutes
• Menu copy for those 6 dishes → 6 minutes
• 30-day menu engineering analysis → 5 minutes
• Quarterly LTO brainstorm → 3 minutes
• Allergen tagging for full menu → 4 minutes

Total: ~26 minutes. The remaining 3.5 hours go to actually cooking and tasting.

A Critical Caveat

AI doesn't know your supplier prices today. A pound of butter costs different things in different markets and at different cases. Always:

• Use your current invoice prices, not AI estimates
• Sanity-check the math (especially yields and conversions)
• Re-cost any recipe with major ingredient swings (proteins, dairy, oil) every quarter

AI does the math. You verify the inputs.

Key Takeaways

• AI can cost recipes, suggest prices, write menu copy, and run menu engineering analyses on demand
• Always feed it your real supplier prices — it doesn't know your market
• Use anti-cliché instructions to keep menu copy on-brand
• Run a Stars/Plowhorses/Puzzles/Dogs analysis every 30 days using your POS export
• Refresh menu copy quarterly with seasonal LTO brainstorms