Decades of digital transformation bypassed corporate travel planning. Fraser explains how a sophisticated multi-agent ecosystem can finally eliminate friction while tightening control.
While modern industries increasingly run on automation, corporate travel remains anchored to decades-old infrastructure and bloated approval chains. Companies are still relying on ignored policy PDFs and reactive expense reports.
Gus Fraser saw an industry paralyzed by its own complexity and built Helix to disrupt it. Moving beyond basic chatbots, Helix utilizes a multi-agent AI ecosystem. These autonomous agents orchestrate budget controls and carbon tracking in real-time, delivering a personalized booking experience without sacrificing financial oversight.
Now stepping in as Staynex’s Chief AI Officer following the Helix integration, Fraser is scaling this AI-native architecture. In this interview, Fraser unpacks the crucial operational leap from chatbots to true agentic AI, and his vision for supercharging business efficiency from the ground up.
Why has corporate travel resisted automation that is seen in almost every other industry?
Gus Fraser: Some of the legacy infrastructure that underpins travel (particularly air travel) is decades old (like GDS). It’s unsuitable for a modern world, let alone an agentic world. There is no universal travel API; it’s very fragmented. This means there is a lot of work involved in coordinating any kind of upgrade, as all parties will need to comply with any new standards.
In corporate travel, particularly, it’s been driven by policy complexity and, traditionally, human knowledge. Companies outsource to travel agents, who have no incentive to disrupt their own margins. Add to this a very low tolerance for error, and there is an overall resistance to change from a technology, people, and risk perspective.
How does a slow and inefficient travel planning process affect businesses and employees?
Gus Fraser: It’s getting easier to book trips manually, but it still wastes hours a skilled employee could be spending on the job they’re being paid for. Not only the traveler themselves, but the approval process can be time-consuming, ineffective, and inconsistent.
How is a line manager, travel admin, or finance admin supposed to evaluate whether the traveler has booked a reasonable flight or hotel? Fixed price hotel budgets don’t work if there is a big conference or event, or if it’s a Tuesday instead of a Friday, or a small town instead of a big city.
These all have a significant impact on price, and it’s ridiculous to expect a finance or even travel admin to have to manually compare prices for every booking on every approval request. The only way is to have policies reflecting real fair prices that are executed in real time, at the time of booking.
How do carbon-tracking mandates shape corporate travel policies?
Gus Fraser: Currently, we track CO2 emissions from every flight taken. In some jurisdictions, such as the EU, there are reporting requirements (like CSRD) that mandate reporting on carbon emissions. Some companies are using separate 3rd party tools to track this, which is obviously inefficient and cannot be accurate.
Eco-conscious organizations can benefit from the transparent reporting to rank flights with lower carbon emissions, or choose to offset, or even take alternative travel (we’re working on integrating rail options).
For local travel, the travel time can be negligible. In the UK, for example, travel time from Edinburgh to London does not change much whether it’s by air or train, but the carbon emissions are significantly different (~122kg CO₂e by air, ~19 kg CO₂e). At scale, this adds up quickly!
What makes agentic AI fundamentally different from chatbots, which may be insufficient to complete complicated tasks?
Gus Fraser: Traditional chatbots are fundamentally request/response services. Ask a question, get a response. An agent can have a task, which could be to find the best flight path, best layover city, or restaurant closest to the hotel that aligns with the traveler’s dietary preferences or requirements.
Helix has over 30 agents, and each is responsible for helping the traveler with options that are just right for that individual. They are hyper-personalized so that they don’t waste time sifting through a wide range of thousands of options, but choose from a few good options.
Agents can maintain state and context across a whole task, not just answer one specific question at a time, so a user could ask for multiple things at once, such as flights, hotel, restaurant, and all actions will be acknowledged and undertaken.
Additionally, agents can be scheduled to run periodically. For example, we always want to know, “What is the most relevant topic to discuss with this traveler right now?” The prompt might be tied to an upcoming trip, like “Remember to check in,” or to changing conditions: “The forecast for next week is 10 degrees colder, so remember to pack a jacket.”
These might seem like simple things, but they significantly improve the traveler experience. Chatbots, by comparison, can only respond to a specific request from a user.
How does Helix consolidate employees and management in travel planning?
Gus Fraser: Helix has a very flexible policy system. A policy is a set of rules that can be customized based on the members of a group. For example, the board might have a different set of rules (regarding cabin class, hotel star rating, etc.) for the sales team. Regardless of role (either employee or management), each user will be in one or more groups, including the “All Employees” group that will involve all users.
Policies are ranked, so that the top policy that applies will be the one that is used. This means that if I am in the “Execs” group and “All Users” group, the Execs group will be ranked higher, and therefore triggered if I am in it, and I conduct a travel search (flights, hotels, etc.).
All of this is seamless to the traveller. Search results presented already take into account the policy that applies to them, whether employee or management. Depending on the organization and policy rules, some travel may be self-approved, or the rules could be configured such that any out-of-policy request must be approved.
By making the policy solution flexible for all organizations, the goal is to facilitate self-booking and self-approval for regular trips, to avoid the inefficiencies and inconsistencies when planning and booking travel manually.
Which features of the Helix platform will take priority during the initial phase of the Staynex integration?
Gus Fraser: Chat-based booking and hyper-personalization using Helix’s memory system to help Staynex travellers (whether corporate or consumer) find better options more quickly. As Helix is already integrated with Sleap.io’s inventory, we will be able to deploy some key features relatively quickly.
I’m particularly interested in rolling out a new user experience paradigm that hasn’t been mastered yet, even by some of the frontier AI companies. At Helix, we call this VOXT — voice and text, not voice OR text (which is often a decision to choose one or the other).
Imagine that you are on a call with a friend discussing a holiday, and your friend says they’ve found a perfect hotel or villa. You then want to see it, so whilst on the call, they send you pictures or links. That is what multimodal should feel like. It should not simply be text-to-speech repetition of what is already in the chat, but a travel companion that actually knows you and can talk to you about recommendations.
How do you plan to scale Helix’s AI-native infrastructure across the entire Staynex organization?
Gus Fraser: As Helix is already integrated with Sleap, it will be a relatively easy lift to apply the AI layer on top of the underlying hotel inventory. The work is ongoing at the moment to integrate the Staynex and Sleap technology stacks, so it won’t be long before Helix AI has been deployed across all public-facing services.
I’m also really interested in rolling out internal agents to support our business functions, as well as the more obvious customer-facing agents we’ve already built. We strongly believe that we can deliver a better service than some of the incumbent online travel agencies, at a fraction of the cost, whilst also saving the traveler time, effort, and money. We can only do this if all my new Staynex colleagues are supercharged with agents.
What will your role be in this merger between Helix and Staynex?
Gus Fraser: As Chief AI Officer of the group, I’ll be not only ensuring that we are leveraging AI to deliver better, faster, and more personalized services to our customer base, but also helping all my colleagues navigate and capitalize on the fast-moving landscape of AI.
The Software Development Lifecycle we developed at Helix had already passed the Turing test; that is to say, we don’t always know if it is a human or an agent that is commenting on or fixing a Github issue or Pull Request! I’m also excited to bring this end-to-end development process to the wider Staynex engineering team so we can do more of the exciting work as humans, and less of the bug finding and testing (which will be delegated to agents).
As an AI-native organization, we need everybody to have an understanding of how AI can help them significantly increase productivity. It is not just incremental gains but orchestrating swarms of agents across all business functions. So there will be a continuous education element as well as building and/or supporting AI tooling for my new colleagues to thrive.
