Recipe Scaling & Yield Conversions
A recipe that works perfectly for 6 portions almost never works perfectly multiplied by 40. Salt does not scale linearly. Reductions do not scale at all (a sauce reducing in a 4-quart pan finishes at a different time than the same sauce in a 20-quart rondeau). Spices need to be pulled back. Garlic needs to be increased. This is one of the hardest things to teach a young cook, and it's one of the highest-leverage things AI can help you with — if you know how to direct it.
What You'll Learn
- The two kinds of scaling AI handles well and the one it does badly
- A scaling prompt template that warns about non-linear ingredients
- How to convert between metric and US units with confidence
- How to use AI for yield analysis (raw to plated weights)
What AI Scales Well
AI scales these reliably:
- Ingredient quantities by weight. Multiply, divide, convert. Trivial for AI.
- Unit conversions. Grams ↔ ounces, ml ↔ cups, °C ↔ °F. AI is essentially always correct.
- Container math. "If my full sheet pan holds 4 quarts of mirepoix at 1/2 inch, how many sheet pans do I need for 18 quarts?" AI handles this easily.
For these, you can trust AI with sanity checks.
What AI Scales Badly Without Help
These need your judgment:
- Salt and seasoning. Salt scales sub-linearly. A recipe that calls for 1% salt at small batch often needs 0.7-0.85% salt at large batch. AI does not know this unless you tell it.
- Aromatic intensity. Garlic, ginger, chili, hard herbs — these often need to be pulled back proportionally as volume increases.
- Reduction times. A sauce reducing by half in a 4-quart pan finishes in 25 minutes; the same volume of liquid reducing in a 20-quart pan with more surface area finishes faster. The same sauce in a 60-quart kettle with less surface-to-volume ratio finishes much slower.
- Cooking times. Roasting a 4 lb pork shoulder is not the same time per pound as roasting a 12 lb shoulder. Conduction works differently at scale.
- Yields. "1 lb of raw spinach yields how much cooked?" depends on water content, season, and your cooking method.
The fix: tell AI what you know it doesn't know. Here's the prompt that captures this.
The Smart Scaling Prompt
Act as my recipe scaling chef.
Original recipe: [paste recipe, with quantities by weight in grams]
Original yield: [X portions of Y grams each]
Target yield: [N portions of Y grams each, OR a total weight]
Scale the recipe and return:
1. A new ingredient list with scaled weights
2. Salt pulled back to 0.85% of total weight (instead of linear scaling)
3. A warning if any ingredient looks suspiciously high after scaling
(especially garlic, fresh herbs, fresh chili, acid)
4. A note on equipment: what size vessel I should be using and how the
reduction or cooking time will likely change at this scale
5. A "verify in test batch first" reminder
Use grams for everything. Show the math for at least one ingredient so I
can sanity check.
The "show the math for at least one ingredient" line is important. It lets you catch the rare case where AI confused 1.5 lb with 1.5 kg in a single conversion.
A Worked Example: Scaling Bolognese for a Banquet
Original recipe (yields 4 quarts, 32 portions at 4 oz):
- Ground beef 80/20: 1.8 kg
- Italian sausage: 0.45 kg
- Mirepoix (onion/carrot/celery): 0.9 kg
- San Marzano tomato: 1.6 kg (2 cans drained)
- Red wine: 360 ml
- Heavy cream: 240 ml
- Olive oil, garlic, herbs, salt: per recipe
Target yield: 240 portions at 4 oz for a wedding.
Run the smart scaling prompt above and you'll get:
- Scaled weights (essentially everything multiplied by 7.5)
- Salt pulled back: instead of 7.5x the original salt, AI recommends targeting 0.8% of total finished weight
- Warnings on garlic (the original 6 cloves becomes 45 cloves if scaled linearly — pull back to 30) and fresh herbs
- Equipment note: 60-quart tilt skillet recommended; reduction will be slower than the home-size batch because of less surface-to-volume ratio — budget an extra 30-40 minutes
- Verify-in-test-batch reminder
Run a 12-portion test batch before banquet day. The whole point of the scaling prompt is to get you to that test batch with a sane starting recipe and a list of things to watch.
Metric ↔ US Conversions
This is where AI shines if you do one thing: always ask in weight, not volume.
A typical American home recipe says "1 cup flour" — but a cup of flour weighs anywhere from 110g to 150g depending on how it's scooped. A professional kitchen runs on weight. When you scale, convert to weight first.
Convert this home recipe to a professional weight-based format:
[paste recipe with volume measurements]
For each ingredient give me:
- Original volume
- Best-estimate weight in grams (for flour, use 120g per cup as standard)
- A note if the conversion is genuinely ambiguous
Then output the recipe in pure-weight format.
Once you've got it in grams, all further scaling math is reliable.
Yield Analysis
Yields — the difference between raw weight bought and plated weight served — are where many food cost calculations break. AI can help you keep a yield table.
Help me build a yield reference table for my kitchen.
For each ingredient give me a typical yield % (edible portion / as-purchased
weight) with a 1-line note on what affects the number:
- Whole salmon (head-on, gutted)
- Whole cauliflower
- Whole pineapple
- Whole chicken
- Raw spinach (cooked yield)
- Mushrooms (sauteed yield)
- Strawberries (hulled yield)
For each, also give me the multiplier I should apply when costing
(1 / yield%).
Save this table. Update it from your own butcher tests. AI gives you a starting reference; your own kitchen gives you the truth.
Key Takeaways
- AI scales weights, units, and container math reliably; it does not handle salt, aromatics, reductions, or cooking times without your help
- Use the smart scaling prompt: pull salt back to 0.85%, warn on suspicious quantities, note equipment changes, remind to test
- Always work in grams (weight), never volume
- Build a yield reference table — AI gives the starting numbers, your butcher tests give the truth
- Always run a test batch before a real-world large-scale service