Using Artificial Intelligence Large Language Model (AI LLM) for Duty of Care Travel Alert

Duty of CareTravel Advisory

Using Artificial Intelligence Large Language Model (AI LLM) for Duty of Care Travel Alert

Duty of CareTravel Advisory

Using Artificial Intelligence Large Language Model (AI LLM) for Duty of Care Travel Alert

Duty of CareTravel Advisory

A key aspect of duty of care provided by travel management companies is offering travel risk advisories to their travelers. This involves continuously monitoring travel-related news or subscribing to specialized travel advisory services.

In this case study, we explore the possibility of using existing Artificial Intelligence Large Language Models (AI LLMs) to automatically interpret news articles and generate conclusions based on their content.

Abstract digital illustration of a network of connected nodes and icons radiating from a central point on a dark blue background, labeled ‘Large Language Model,’ representing AI data processing.Abstract digital illustration of a network of connected nodes and icons radiating from a central point on a dark blue background, labeled ‘Large Language Model,’ representing AI data processing.Abstract digital illustration of a network of connected nodes and icons radiating from a central point on a dark blue background, labeled ‘Large Language Model,’ representing AI data processing.

For this example, we will use a news release from the Ministry of Foreign Affairs of Malaysia.

TRAVEL ADVISORY FOR PATTANI, YALA AND NARATHIWAT

The Ministry of Foreign Affairs, through the Consulate General of Malaysia in Songkhla, is closely monitoring the situation in the provinces of Pattani, Yala and Narathiwat following recent security incidents.
 
In light of these developments, Malaysians travelling to or residing in these provinces are advised to exercise caution and remain vigilant. The Ministry also recommends postponing all non-essential visits to the provinces for the time being.
 
As a precautionary measure, Malaysians currently in these provinces are urged to register their presence with the Consulate General of Malaysia in Songkhla.

We conducted tests using ChatGPT, Gemini, and DeepSeek by supplying them with the article text and instructing the AI to generate the following summary:
Dark-themed interface displaying a structured travel alert summary with fields for travel alert status, category, severity level, affected country, provinces, and airport code.Dark-themed interface displaying a structured travel alert summary with fields for travel alert status, category, severity level, affected country, provinces, and airport code.Dark-themed interface displaying a structured travel alert summary with fields for travel alert status, category, severity level, affected country, provinces, and airport code.

Test result using the example above.

ChatGPT
Screenshot of a ChatGPT interface summarizing a travel advisory for Malaysians in Thailand, followed by a structured output listing alert status, category, severity level, and affected locationsScreenshot of a ChatGPT interface summarizing a travel advisory for Malaysians in Thailand, followed by a structured output listing alert status, category, severity level, and affected locationsScreenshot of a ChatGPT interface summarizing a travel advisory for Malaysians in Thailand, followed by a structured output listing alert status, category, severity level, and affected locations

Gemini
Screenshot of a Gemini AI interface summarizing a Thailand travel advisory, with a formatted travel alert output including alert status, category, severity level, and affected provinces.Screenshot of a Gemini AI interface summarizing a Thailand travel advisory, with a formatted travel alert output including alert status, category, severity level, and affected provinces.Screenshot of a Gemini AI interface summarizing a Thailand travel advisory, with a formatted travel alert output including alert status, category, severity level, and affected provinces.

Deepseek
Screenshot of a DeepSeek interface displaying a formatted travel alert summary for Malaysians traveling to Thailand, including structured fields for severity level and affected areas.Screenshot of a DeepSeek interface displaying a formatted travel alert summary for Malaysians traveling to Thailand, including structured fields for severity level and affected areas.Screenshot of a DeepSeek interface displaying a formatted travel alert summary for Malaysians traveling to Thailand, including structured fields for severity level and affected areas.

Based on the test results, all three—ChatGPT, Gemini, and DeepSeek—were able to generate contextual summaries. However, ChatGPT demonstrated a stronger ability to connect with relevant external context, such as identifying the appropriate IATA airport codes.

As a result, we will continue the evaluation using only ChatGPT to determine its capability in interpreting articles from general news sources and whether it can provide more in-depth external data, such as the geolocation of affected areas, in a machine-readable format.
Dark-themed interface showing instructions to analyze an article and format a travel alert summary into JSON, including fields for alert status, category, severity level, affected country, affected areas with geolocation, and airport code.Dark-themed interface showing instructions to analyze an article and format a travel alert summary into JSON, including fields for alert status, category, severity level, affected country, affected areas with geolocation, and airport code.Dark-themed interface showing instructions to analyze an article and format a travel alert summary into JSON, including fields for alert status, category, severity level, affected country, affected areas with geolocation, and airport code.

Json
Screenshot of a JSON-formatted travel alert indicating a political advisory for Thailand with second-level travel severity, affected cities Pattani, Yala, and Narathiwat, including latitude and longitude coordinates and airport code HDYScreenshot of a JSON-formatted travel alert indicating a political advisory for Thailand with second-level travel severity, affected cities Pattani, Yala, and Narathiwat, including latitude and longitude coordinates and airport code HDYScreenshot of a JSON-formatted travel alert indicating a political advisory for Thailand with second-level travel severity, affected cities Pattani, Yala, and Narathiwat, including latitude and longitude coordinates and airport code HDY

Tested with negative case - articles not of travel advisory context

Online news article reporting political developments in Singapore, discussing opposition party strategies and parliamentary changes ahead of elections.Online news article reporting political developments in Singapore, discussing opposition party strategies and parliamentary changes ahead of elections.Online news article reporting political developments in Singapore, discussing opposition party strategies and parliamentary changes ahead of elections.

Json
Screenshot of a JSON response showing no travel alert triggered, with null values for alert category, severity level, affected country, areas, and airport code.Screenshot of a JSON response showing no travel alert triggered, with null values for alert category, severity level, affected country, areas, and airport code.Screenshot of a JSON response showing no travel alert triggered, with null values for alert category, severity level, affected country, areas, and airport code.

Tested with possible advisory context in Channel News Asia

Online news article describing heavy snowfall and cold weather conditions in northern China, including flight cancellations and transportation disruptions in Beijing.Online news article describing heavy snowfall and cold weather conditions in northern China, including flight cancellations and transportation disruptions in Beijing.Online news article describing heavy snowfall and cold weather conditions in northern China, including flight cancellations and transportation disruptions in Beijing.

Json
Screenshot of a JSON-formatted travel alert for China categorized as a natural disaster with second-level travel severity, listing Beijing and Inner Mongolia with geolocation data and airport code PEK.Screenshot of a JSON-formatted travel alert for China categorized as a natural disaster with second-level travel severity, listing Beijing and Inner Mongolia with geolocation data and airport code PEK.Screenshot of a JSON-formatted travel alert for China categorized as a natural disaster with second-level travel severity, listing Beijing and Inner Mongolia with geolocation data and airport code PEK.
This is an initial exploration of AI capabilities, and additional test cases across multiple dimensions are still required.

 

That said, the process can likely be automated—by crawling news websites, submitting relevant questions to an AI model, and capturing the structured output directly into our database.
Using Google AI Model to Retrieve Carbon Emission for Flights
Corporate Booking ToolOnline Booking SiteEnvironmental Responsibilities
Using Google AI Model to Retrieve Carbon Emission for Flights
Corporate Booking ToolOnline Booking SiteEnvironmental Responsibilities
Using Google AI Model to Retrieve Carbon Emission for Flights
Corporate Booking ToolOnline Booking SiteEnvironmental Responsibilities