Learning module
Practice this topic →Read the concept, complete the knowledge checks, then practice the same skill in Coffee Lab.
AI Barista
01 · Translate natural language into menu decisions.
Translate natural language into menu decisions.
AI Barista is not just a chatbot. The simulation maps customer language into structured preference dimensions, ranks menu profiles, respects exclusions and suggests recipe customizations.
02 · Preference dimensions
Preference dimensions
StrengthGentle → balanced → bold.
SweetnessNo added → light → regular → indulgent.
CreaminessClean → smooth → extra creamy.
FreshnessComforting → refreshing / citrus-forward.
TemperatureHot, iced or flexible.
FlavorCaramel, hazelnut, floral, chocolate, berry, lemon, matcha, palm sugar and more.
MoodFocus, energy, calm, comfort, social, curious, indulgent, adventurous.
ContextCommute, meeting, breakfast, after lunch, study, weekend and takeaway.
03 · Full signal model used by the enriched AI Barista
Full signal model used by the enriched AI Barista
| Signal | Examples | How to treat it |
|---|---|---|
| Coffee requirement | Coffee / no coffee | Hard category filter |
| Strength | Gentle, medium, strong, extra-bold | Menu match + possible shot adjustment |
| Sweetness | No added, 25%, half, regular, sweeter | Scale base sweetener/syrup |
| Texture | Clean, smooth, creamy, extra foam | Menu fit + milk/foam customization |
| Freshness | Comforting vs refreshing | Separates coffee/dessert families from refreshers |
| Acidity / bitterness | Low acidity, not bitter | Secondary sensory constraints |
| Temperature | Hot / iced / flexible | Service filter |
| Flavor likes | Palm sugar, caramel, hazelnut, floral, chocolate, berry, lemon, matcha, etc. | Positive ranking signal |
| Flavor exclusions | No caramel, avoid rose, etc. | Hard negative signal |
| Mood | Focus, calm, comfort, social, adventurous | Soft ranking signal |
| Occasion | Commute, meeting, breakfast, after lunch, study, weekend | Soft context signal |
| Milk | Dairy, oat, soy, lactose-free | Customization / service constraint |
| Size & foam | Short/regular/large; none/light/extra foam | Recipe modifiers |
| Adventure level | Classic vs surprise me | Controls how far ranking can move toward experimental signatures |
04 · Five-step recommendation method
Five-step recommendation method
Extract hard constraints
Examples: no coffee, hot only, oat milk, no caramel.
Extract sensory targets
Strength, sweetness, creaminess, freshness, bitterness and acidity.
Identify mood and occasion
Use them as secondary signals, not stronger than explicit constraints.
Rank menu fits
Prefer a naturally matching drink over excessive modifications.
Explain and customize
Tell the guest why the recommendation fits and what you would change.
Example: “Strong, iced, oat milk, half sweet, not bitter, no caramel, for a commute” should trigger a strong coffee candidate, enforce iced/oat/no-caramel constraints, reduce sweetness, then check bitterness and context fit.
What should the AI Barista prioritize first?
Practice Lab · Intelligent Barista
Launch Simulation →AI Barista Scenario Lab
Interpret nuanced customer prompts and test whether the AI recommendation fits mood, flavor, strength and dietary preferences.