Agentic Voice Assistant for Airport Operations

Solution Overview

Airport ground operations support over 40 million flights annually and involve numerous tasks like baggage handling, refueling, catering, and maintenance. This complexity is compounded by less-experienced staff, which increases the likelihood of human error, threatens safety, and leads to costly flight delays. For instance, a 15-minute delay for an Airbus A321 can cost airlines around €3,030 (approximately $3500).

This solution uses Dataworkz, Google Cloud and MongoDB Atlas to power an agentic voice assistant for airport ground operations. This assistant guides operators through essential checklists, retrieves real-time answers from embedded manuals via Dataworkz’s RAG application, and logs actions for audit and optimization. The system enables hands-free interaction by using Vertex AI for speech-to-text, text-to-speech, and NLP, and uses Atlas Vector Search for context-aware retrieval. Staff can use the voice assistant as an intelligent knowledge base, which ensures compliance, increases operational efficiency, and reduces the training time for ground crews.

Reference Architectures

This solution has the following architecture:

Figure 1. Intelligent airport ground operations architecture

As shown in Figure 1, this solution uses the following technologies:

• Dataworkz for RAG and agentic AI: Dataworkz is a managed RAG platform that provides an agentic AI framework and an AI-optimized data layer. It facilitates seamless data integration into MongoDB Atlas and manages the end-to-end RAG workflow by ingesting and embedding technical manuals and regulations with Voyage AI models. Dataworkz's RAG builder integrates the voice assistant with complex information from safety manuals that it stores in MongoDB. After the end-to-end workflow completes, Dataworkz then executes queries with MongoDB Atlas Vector Search for accurate information retrieval.

• MongoDB Atlas as the core database: MongoDB Atlas serves as the application's centralized data store. Its flexibility and scalability allows it to store large volumes of unstructured data that are common in AI workloads. Additionally, its aggregation framework simplifies data processing for AI application development. Atlas also stores session activity logs and raw and vectorized text data from embedded manuals, which Dataworkz queries using Vector Search.

• Google Cloud for speech and NLP: Vertex AI from Google Cloud provides essential speech-to-text, text-to-speech and NLP capabilities, enabling natural interactions with the agentic voice assistant for ground crew. For example, if an operator asks an aircraft question, such as What do you mean by nose gear torsion? Vertex AI performs speech-to-text conversion and sends the information to Dataworkz, which intelligently queries MongoDB using Vector Search.

• Voice assistant functionality: The agentic voice assistant dynamically feeds checklist items to operators. The system provides instant validation, reads next steps, and offers additional context from manuals if it detects discrepancies. Dataworkz enables intelligent querying of embedded manuals via MongoDB Atlas Vector Search, delivering context-specific answers in real-time.

Data Model Approach

MongoDB's flexible document model supports the unstructured data used in AI development, such as data from complex monitoring and logging systems in the airport environment. MongoDB Atlas stores agent memory, including the current state of checklists, individual steps, and RAG-generated logs. These logs help track queries for training, and audit trails for security compliance.

The code snippet below shows how the logs collection records NLP queries.

{
   "_id": {
      "$oid": "686ffc11d66babf8cd958229"
   },
   "sessionId": "efc07eac-ef36-4487-adc0-5c0f82eddfe7",
   "logs": [
      {
         "timestamp": "2025-07-10T17:44:48.838Z",
         "toolName": "consultManual",
         "type": "call",
         "details": {
            "name": "consultManual",
            "args": {
               "query": "What is a tow bar operation?"
            }
         }
      }
   ]
}

The solution logs and stores each ground operation as a JSON document in MongoDB Atlas for audit and optimization. Each document has a unique sessionId, which you can use to update the documents as necessary.

Build the Solution

You can replicate this solution by following the README in the GitHub repository, which guides you through the following steps.

git clone https://github.com/mongodb-industry-solutions/aircraft-groundops-sync.git

npm install

npm run dev

npm run build

Key Learnings

• Reduction in training time and cognitive load: With this solution, ground crews do not need to memorize every detail of extensive manuals. The voice assistant acts as an intelligent knowledge base that delivers context-aware answers in real-time, and reduces the cognitive burden of keeping track of the operation steps and aircraft manuals.

• Enhanced error prevention and compliance: The compliance checklist provides immediate access to correct procedures, minimizes the risks of human error due to forgotten steps, and enhances safety and compliance through session activity logs.

• Data-driven optimization: Analyze and audit your data by using session activity logs stored in MongoDB Atlas. These logs enable continuous optimization by providing data-driven insights which identify areas that need additional training or improvements in manual content.

• Flexible and scalable AI-powered guidance: MongoDB’s flexible document model, real-time processing capabilities, and advanced vector search can be integrated to build an agentic voice assistance system. This architecture helps AI applications handle complex operational environments by seamlessly integrating structured and unstructured data.

Authors

• Romina Carranza, MongoDB

• Humza Akthar, MongoDB

Learn More

• Agentic AI-Powered Connected Fleet Incident Advisor

• Elevate Flight Operations with Real-Time Analytics

• Transforming the Driver Experience with MongoDB & Google Cloud

• To learn how MongoDB supports manufacturing and automotive applications, visit MongoDB for Manufacturing & Mobility.