What is Procurement asking about AI?
Procurement and Travel Managers naturally ask many questions about AI to their travel management company (TMC) and in the RFP process. Here are some of the most common questions about AI and travel programmes, and what you should look for in a supplier’s answer.
This chapter is taken from Artificial intelligence, simplified. A corporate travel manager’s practical guide to implementing AI. Get your copy here.
How do we know that travellers get the right answers with our TMC’s AI?
The AI should allow users to upvote or downvote answers. This way, the LLM will learn and refine its responses, and your TMC's AI team can use these ratings to understand where the model made an error and establish guardrails for future answers.
Look out for how the TMCs will ensure that tools align to your specific business travel policy and travel programme requirements.
FCM’s answer: Agentic AI is integrated into every FCM touch point and tailored to your travel policy. Working with our team, you’ll build a playbook containing your travel policy and requirements, which will then be used to train our agentic AI, known as Sam. This means more personalised outcomes and reduces the risk of hallucinations.
Are our travellers’ interactions being used to train the AI?
It’s important to distinguish between the model and the tool.
The AI base model, e.g. ChatGPT, Gemini, or Claude behind the tool, should be a fully trained framework ready for deployment in your programme.
The AI tool you use should only store interactions and learning data within itself, and not push them into the wider base model.
FCM’s answer: Yes, each interaction with our agentic AI, Sam, helps build its knowledge and create better outputs. However, those interactions are securely stored in FCM's Sam ecosystem. Sam is built on Anthropic’s model, but interactions and data are not fed into that model.
How does the TMC’s AI integrate with existing AI tools?
This will vary from supplier to supplier. Some solutions may not be able to be integrated into another for various reasons.
Your question could be expanded to include integration across the travel management company (TMC); how employees use it, how it’s speeding up tasks, and what travel consultants and account managers can access.
FCM’s answer: As it is tailored to your travel programme and available exclusively in FCM Platform, FCM’s agentic AI will only work with the playbook and data available in the platform and interactions.
As Sam is built on Anthropic’s model, it has access to the internet for queries related to the weather for a trip, or the latest news from an airline or airport.
Where Sam is unique is how it’s embedded across the entire FCM business, meaning that travel consultants and account managers have access to the same insights and intelligence as your employees. This integration allows for quicker decision-making, streamlined processes, and a seamless experience for all users.
How does a TMC ensure that AI won't hallucinate and give the wrong answers?
The refinement of the base LLM model is key here. Suppliers should report on model accuracy and explain how guardrails are put in place to keep answers accurate.
FCM’s answer: This is where building a customised playbook is crucial. With your specific requirements and instructions in one place, it reduces the risk of hallucinations.
Artificial intelligence, simplified.
A corporate travel manager’s practical guide to implementing AI
Download the full guide for:
- A brief history of AI
- How to implement successful AI pilots
- How to avoid failure
- Reviewing TMC AI readiness, including a checklist