NVIDIA AI Enters Navan’s Travel Operations in a Major Push to Modernise Business Trips in San Francisco, USA

Image generated with Ai

Business travellers could soon spend less time dealing with disrupted journeys and booking changes as San Francisco, USA-based travel technology company Navan expands its use of artificial intelligence (AI). On 16 October 2026, Navan announced that it had deployed NVIDIA Nemotron 3.5 Super VL to power several of its AI travel agents in production. The move aims to make digital travel assistance more responsive, efficient and capable of handling complex requests. Source: Navan

Navan Brings a New AI Model Into Travel Operations

Navan, which provides business travel booking and expense management services, says its latest deployment was completed within three weeks of receiving early access to the NVIDIA model. Rather than using the technology in its original form, the company adapted it for travel-related tasks.

The deployment forms part of Navan’s strategy of combining open and closed AI models. It selects models for specific tasks, fine-tunes them with travel expertise and hosts selected systems on its own infrastructure. According to the company, this approach gives it greater control over response speed, operating costs and reliability.

For companies arranging frequent employee trips, improvements in these areas could make travel management less complicated. However, Navan has not announced that the new model is available to every traveller or that it guarantees faster resolution of every booking problem.

Flight Disruptions Are a Key Test for Travel AI

A cancelled flight can create a chain of problems for business travellers. Passengers may need to find another departure, change hotel bookings, adjust ground transport and inform colleagues about a delayed arrival.

Navan says an effective AI travel agent must do more than explain that a flight has been cancelled. Its system needs to identify alternatives, present the available choices and help the traveller complete the necessary changes.

The platform uses specialised AI agents for different parts of a journey, including flights, hotels, cars and trains. A coordinating agent can direct a request to the relevant specialist while allowing information to be shared across the trip.

This approach could make it easier to manage connected travel arrangements through one service. The practical benefit will depend on the availability of suitable alternatives, access to live travel information and the system’s ability to complete the required transactions.

NVIDIA Model Shows Promising Results in Company Tests

NVIDIA released Nemotron 3.5 Super VL on 15 October 2026. The model is designed for AI systems that need to reason through multiple steps, use digital tools and process different forms of information.

Navan evaluated the model against its existing self-hosted production model and other competing systems. The company used consistent prompts, travel scenarios, tools and evaluation criteria before deciding to deploy the fine-tuned version.

According to Navan, the adapted model beat or matched competing models on six of seven evaluation measures. These included tool use, task completion, following constraints, using relevant information and maintaining useful conversations. The company also reported the lowest latency in its comparison.

These findings are based on Navan’s internal evaluation rather than an independent industry-wide assessment. They indicate that the model performed well in the company’s tests, but they do not establish a specific improvement in customer satisfaction or prove that every real-world travel request will be completed successfully.

Fine-Tuning Helps Adapt AI to Travel Needs

General-purpose AI models can understand language, but they do not automatically know how a particular travel platform handles bookings, company policies or support requests.

Navan used NVIDIA’s NeMo AutoModel training tools and a technique known as low-rank adaptation (LoRA) to customise the model. This method allows developers to adapt selected model parameters without retraining the entire system from scratch.

In its technical explanation, Navan says its final training set contained around 10 million tokens of supervised training information, alongside nearly 2,000 validation tests. The company focused its evaluation on text-based travel tasks and tool use.

The model has approximately 120 billion total parameters but activates around 12 billion for each processing step, according to Navan. This mixture-of-experts design is intended to balance the model’s reasoning capacity with speed and computational efficiency.

Potential Benefits for Corporate Travel Management

Faster, more capable AI assistance could help travel management teams handle routine enquiries while allowing human professionals to concentrate on complicated cases. It may also help employees navigate approved travel options and company spending rules.

Navan says the deployed model is already powering customer conversations at a lower cost. However, it has not disclosed a precise percentage saving, and the announcement does not establish how much individual companies will save on their travel programmes.

Privacy and accountability will remain important as AI systems handle traveller information, itineraries and business-related requests. The US Government Accountability Office highlighted privacy risks associated with AI in a March 2026 report, while the Federal Trade Commission has emphasised that AI companies must honour their privacy and confidentiality commitments. These broader government findings do not indicate any specific wrongdoing by Navan.

AI Could Reshape the Business Travel Experience

For travellers, the real test of this technology will not be how impressive an AI conversation sounds. It will be whether the system can help them reach their destination when plans change, explain their options clearly and reduce the effort involved in managing a trip.

Navan’s deployment marks another step towards travel services that can act on requests rather than simply provide information. Its internal results offer an encouraging sign, although independent evidence of improved traveller outcomes is still needed. As travel companies continue to adopt AI, reliability, data protection and access to human support will remain essential to building trust.