A new article by Assoc. Prof. Murat Doğan—Vice Dean of the Faculty of Fine Arts at Istanbul Gelisim University (IGU) and a faculty member of the Department of Gastronomy and Culinary Arts—has been published in Journal of Hotel Restaurant & Hi-Tech. The article appears under the title "Prompt Engineering and a Future Vision in Restaurant Management". The full text of the article is presented below.
Dear readers, for some time now, my monthly columns have focused on innovative approaches applicable in the food and beverage industry. In this month’s article, I will discuss prompt engineering—an area still very new in global practice and, as far as I have observed, not yet applied in Turkey’s restaurant industry.
Together with my PhD student, we have spent considerable time pondering the question: “How can prompt engineering be applied in the restaurant sector?” We presented our scientific findings as a paper at an international congress in February. In this article, I will try to explain the subject in a simplified way.
What is Prompt Engineering?
It’s useful to begin with the word “prompt” itself. While we may not have heard the term “prompt” frequently in past years, we’ve all likely encountered the device known as a prompter and the occasional mishaps associated with it. As you know, a prompter is essentially an electronic teleprompter. From this, we can easily infer that at the root of the word “prompt” lies the idea of assisting or guiding. Since the early 15th century, in Western languages, “prompt” has meant “to encourage action” or “to assist a speaker with lines.” Today, in fields dominated by artificial intelligence, “prompt” refers to the command or input given to an AI system to perform a specific task or generate content.
Since the release of ChatGPT, there has been significant advancement in areas like natural language processing, code generation, and autonomous systems. However, the reasoning capabilities of the Large Language Models (LLMs) these programs use remain limited. Research has shown that LLM agents, which are adaptable to a broader range of applications through the integration of various tools, can be developed. In these new models, prompts play a critical role. Prompts serve as directives that guide LLM outputs and help automate processes.
How Prompt Engineering Emerged
Prompt engineering is a complex and time-consuming process. To address this, a method called Automatic Prompt Engineering (APE) has been developed, enabling LLMs to generate their own prompts. APE allows for the creation of more effective and robust prompts with fewer attempts. Additionally, certain strategies and textual structures have been established to ensure that LLMs produce desired outputs and avoid common errors.
Applicability of Prompt Engineering in Restaurant Operations
In restaurant management, prompt templates can serve a function similar to software patterns—facilitating knowledge transfer and offering reusable solutions to common problems. These templates consist of elements such as instruction, input, role, output format, tone/style, and constraints, which I will briefly explain below.
Instruction: Specifies what the model is expected to do.
Input: Provides the necessary context for the response.
Role: Shapes the model’s behavior and perspective.
Output Format: Defines the structure and format of the response.
Tone/Style: Determines the tone and level of formality in communication.
Constraints: Ensures responses adhere to ethical and technical boundaries.
These elements ensure that the data obtained is useful and comprehensible. In 2023, Jules White and colleagues classified these prompt elements in their work titled “A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT” into “Main Categories and Patterns”. These patterns serve as valuable tools for effective communication and problem-solving in restaurant operations.
Prompt Engineering: Key Categories and Patterns for Restaurants
Various techniques and strategies related to AI usage in restaurant management aim to ensure more effective and efficient utilization of AI. Main headings and examples include:
-
Input Semantics:
-
Output Customization:
-
Persona: The AI adopts a specific role. Example: Acting like a chef to suggest a new menu item.
-
Template: AI provides output in a predefined format. Example: Using a standard format to introduce a new dish.
-
Visualization Generator: AI produces visual outputs related to topics like menus or restaurant layouts. Example: Creating a diagram for table arrangements.
-
Error Identification:
-
Prompt Improvement:
-
Question Refinement: Generic customer queries are transformed into more specific ones. Example: “What do you recommend?” becomes “What are today’s specials?”
-
Alternative Approaches: AI offers different solutions to a problem. Example: Suggesting alternative ingredients to make a dish vegan.
-
Interaction:
-
Flipped Interaction: AI collects information from customers or staff. Example: Asking questions to gather data for creating a new menu.
-
Game Play: Adds a fun, interactive element for customers. Example: A guessing game about dish ingredients.
-
Context Control:
These techniques aim to enhance customer experience and optimize operational processes by enabling AI to be used more effectively in restaurant management.
Menu Management and Personalization with Prompt Engineering
Menu management consists of core topics such as menu pricing, menu design, menu operations, and menu analysis, which I briefly outline below.
-
Menu Pricing:
-
Pricing directly influences customer perception and sales. Example: Prices ending in “9” can enhance the perceived value.
-
“Decoy” items can guide customer pricing perception.
-
While cost and profit goals are essential, customer expectations must also be considered.
-
Menu Design:
-
The menu is a reflection of the restaurant’s brand. Design elements like color, fonts, and visuals influence customer decisions.
-
“Sweet spots” (areas the eye is drawn to first) can increase item sales.
-
Names and descriptions of items affect customer choices.
-
Menu Operations:
-
Food safety and hygiene are particularly critical for small businesses. Example: HACCP-based risk systems are vital.
-
Standardized recipes, portion control, and accurate demand forecasting help lower costs and improve satisfaction.
-
Menu Analysis:
-
Menu items are evaluated based on metrics such as cost, popularity, and profitability.
-
Strategies like changes in presentation, recipe, or pricing can be applied to low-performing items.
-
Multi-dimensional analysis methods have been developed to address gaps in traditional evaluations.
-
Chatbot-Driven Menu Personalization:
-
AI and chatbots can be integrated into menu creation and personalization. Example: Models like ChatGPT can be fine-tuned to contribute to menu development.
-
These technologies can effectively create personalized menus tailored to customer preferences.
In conclusion, restaurant menu management is a complex process that involves many factors such as pricing, design, operations, and analysis. Prompt Engineering can make these processes more efficient and customer-centric.
Stay well,
Assoc. Prof. Murat Doğan
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
