How to create a custom GPT for a travel agency technician

Generative artificial intelligence has become a key tool for travel agency technical teams. But for it to truly add value, it's not enough to "ask things" to a generic chat. It is necessary to design a personalized GPT, adapted to the agency's daily operations, with professional judgment, clear limits, and the capacity to work with both external sources and proprietary documentation.

In this article, we explain how to create a GPT designed for a travel agency technician, in a generic and reusable way, applicable to multiple areas: operations, product, suppliers, regulations, internal processes, or support for the sales team.

STEP 1. Define the GPT's objective and scope

The first step is to specify the role the GPT should play within the organization. Some common use examples are:

• Support for operational and regulatory information retrieval
• Analysis of documentation from suppliers or official bodies
• Generation of checklists and internal procedures
• Help in drafting coherent responses for clients and teams
• Centralization of dispersed knowledge

It is important to define what it can and cannot do: the GPT must be an assistant, not a substitute for professional judgment or a definitive source.

STEP 2. Design a "structured search" mode with reliable sources

A useful GPT must know how to search for information with judgment. It's not about finding any answer, but about:

• Understanding the context of the query (type of service, country, dates, exceptions).
• Prioritizing official and reliable sources, such as:

• Public bodies
• Official suppliers
• Recognized sectoral entities
• Internal agency documentation

• Avoiding speculation or unverified information.
• Properly documenting the sources consulted.

This search pattern is applicable to many areas: regulations, commercial conditions, operational processes, or specific service requirements.

STEP 3. Add the capacity for analyzing documentation provided by the user

One of the great advantages of a personalized GPT is that it can work with the agency's real documents and not just documents from the internet (which it can also do).

Common document types

• Official PDFs
• Supplier circulars
• Internal manuals
• Presentations, infographics, or images

What can the GPT do with these documents?

• Summarize relevant information
• Detect critical points, dates, and exceptions
• Compare criteria between documents
• Convert the content into:

• Operational checklists
• Summaries for teams
• Reusable practical guides

This allows for standardizing knowledge and reducing dependence on specific individuals.

STEP 4. Create the main prompt: recommended structure

The prompt is the core of your GPT. It must be clear, structured, and explicit. A good practice is to divide it into blocks:

1. Main capabilities and functions

• Structured search with reliable sources
• Analysis of provided documentation
• Clear synthesis oriented towards agency operations

2. Limits and ethical boundaries

• Do not give definitive legal, fiscal, or regulatory advice; contrast the answer with an expert's experience.
• Always recommend verification with official sources
• Do not invent information if it is not truly documented

3. Format requirements

• Numbered lists
• Tables when necessary
• Executive summaries
• Final sources section

4. Establish limitations

• Information may change
• If context is missing, it must ask for it before answering

5. Tone and style

• Professional
• Clear
• Practical
• Oriented towards agency technicians and teams

6. Security

Define critical response instructions from the start, such as:

STEP 5. Create your personalized assistant

5.1 Create your GPT on Chat-GPT

Practical steps:

• Access the Create a GPT option.
• Paste the structured prompt.
• Add examples of common questions.
• Test it with real cases and real documents.
• Adjust behavior and format.

To publish and share it publicly, you need ChatGPT Plus, Pro, or Enterprise, and to have the creator profile enabled.

5.2 Create a Gem in Gemini

The process to do the same is similar:

• Create a Gem from the Gemini manager.
• Define instructions and behavior.
• Test it with real cases and real documents.
• Save and share it according to account permissions (especially in Workspace environments).

Gemini is a good option if you already work regularly with Google Docs, Drive, and Workspace.

STEP 6. Testing, documentation, and internal adoption

Before deploying it to the team:

• Test real cases and edge cases
• Verify that it respects security limits
• Ensure it always asks for context when necessary
• Create a list of advanced users to test the GPT before putting it into production

Finally, create a brief internal user manual: what it is for, how to ask it questions, and how to interpret the answers.

Conclusion

A personalized GPT for a travel agency is a strategic tool for supporting knowledge and operations. If well designed, it allows for gaining efficiency, reducing errors, and sharing judgment within the organization. The key is not the technology, but defining the steps, limits, and the professional use it will be put to.