FCM, the corporate travel management company (TMC), has taken a deliberate approach to artificial intelligence (AI)—building an in-house platform rather than bolting on off-the-shelf tools. The result is an internal system which powers its AI assistant Sam alongside in-house processes.
In a Q&A, John Morhous, global chief experience officer of Flight Centre Travel Group’s corporate brands, explained the decision, early successes and common AI misconceptions.
What made you decide to build versus buy in this AI context?
We're certainly using third-party frontier large language models (LLM) as the horsepower for sharing information and using generative AI to produce outcomes. Our enterprise relationship is with Anthropic.
But the “Mosaic” platform (we don’t use the name externally) is an investment we've made because we feel that in the corporate enterprise space, customers need more safety and security in how they’re using AI.
Anybody can easily slap a ChatGPT chatbot onto the experience and tick a box that they're doing AI. What we built in the middle is a controlling guardrail system, which is key. If you look at AI from an agentic environment, to get to an agentic model, you have to give your LLM access to tools and skills. The tools are what give us access to specific data such as travel policy information, profile information, flight or hotel availability. You need to give it tools that say go accomplish this objective. So we have built in our platform tools to access our internal systems and trusted third-party platforms such as CIBT that customers can opt in or opt out of depending on their program.
The last piece is what we call skills playbooks. That allows customers to build guardrails around how the AI is to work with the tools that we provide them. A ChatGPT bot could provide an answer that is the summary of the internet on whether a visa is needed for Australia. The answer could be correct in many situations, but what the customer does not have is control. So, Sam is built on this guardrail system and these playbooks we have created. It gives companies more comfort that when they’re enabling these LLM services in their travel programs, the answers will be what they want and not the collective wisdom of the internet.
How are you thinking about your AI roadmap and how are you prioritizing?
I have a dedicated team that just works on AI-based solutions that has ownership over their own product offering—that entire playbook model is essentially a product offering they have for our internal businesses and businesses then build on top of the capability that they want. Our approach is to have the AI team just be another product team within our business that creates capabilities that our business then uses for those different services. We've built this as a product shop within our ecosystem and then the product teams work with either external customers or internal customers using the same platform stack.
How do you decide what's next? It's easy to get overexcited about AI, how do you stop that happening?
It's something we spend a lot of time trying to figure out strategically. From an overarching perspective, we’ve embedded into both brands this concept of what we call design and control. We ultimately want to be designing and controlling the full end-to-end experience as best we can with our customers in either brand.
It's not like I have an AI roadmap per se, what we have is product roadmaps that include some AI capabilities. I'm a strong believer that AI is just a capability that we're enabling through our tools, that you need to leverage. And, just like any good capability, we need to figure out how it impacts the function that we're doing, what value creation we're doing with the specific product or service offering that we have. How can that service offering be improved? It's my job to then look at and try to read the tea leaves as to what's going on with the macro technology world and try to understand where things are happening. It’s very hard right now because things are moving so quickly. What’s important is not losing sight of where you’re creating value.
Recent Phocuswright research revealed most companies haven't rewired for AI yet and don’t have it as core infrastructure. Where are you?
We've had our AI team assembled for the past 18 months. We have a good portion of initiatives using AI tools that are producing solid documented ROI so I don't think we’re at the back of the pack. Keep in mind that the role we provide is to facilitate people getting together in person to do something. Current workflows to do that require processes that have nothing to do with what you're trying to achieve and AI tools can help with that.
We have a good portion of initiatives using AI tools that are producing solid, documented ROI so I don't think we’re at the back of the pack.
John Morhous, FCM
I've got customers that want to have 95% online self-service type of transactions. Others are completely offline because they have a huge trust factor and large cost and want a person to go through and be sure that their flight is going to make the connection. Neither is right. It's about establishing that connection point between people. So we're not throwing the baby out with the bathwater. We are just trying to find ways to optimize what we do using the technology that's available.
Your 2025 hype cycle assessment for business travel placed us at the peak of inflated expectations. Where are we now and how quickly might we move forward?
I certainly think we're near the top, probably coming back down the backside of it a little bit more. I don't think we're at the valley of complete desolation yet. If you go into an AI tool and you ask it to give you an assessment of news headlines for the last 12 months you would see a period where pretty much everything in the world was gonna be AI-ified in some way. Now, I think you're starting to see more pragmatism. It's not like every company is folding because everything has been replaced by bots. You're seeing more of a gentle push by company leaders saying this is no longer a nice to have, it's something we're expecting you to do.
Once you get past the hype, that top where it's so outlandish, you get into that component where you're starting to see some interesting things. We're getting to a really interesting time where we’re going to see a lot more value be extracted from it. It’s not going to be these systems that do everything automatically, it's going to be incremental.
What are one or two common misconceptions around AI in the corporate travel world?
People expect everything to be AI-ified automatically and suddenly it's infinitely better is a common misconception. It's a tool that requires a little bit of experimentation, a lot of curiosity and some guardrails and controls. You can get a lot of benefit from it if you know how to work with it, but you have to invest your own time to do it.
The second misconception, I think, is there is a lot of fear that comes naturally from people when adapting to change. We often misconstrue lack of progress with people not moving past their fears because they don’t know how it works. And this is one of those situations where you're probably not going to know exactly how it works. You may have a rough idea, you just have to work with it to understand how the outputs come. And it's non-deterministic so you're not sure what you're going to get out of it and that gives people anxiety. My message is to get past that and embrace what it can do and just work with it and with what it's good at.