The rise of ChatGPT shows just what AI can do with large amounts of data. When the GPT-3 large language model was released in 2020, it was revealed that it had been trained on more than 430 billion tokens– word and phrases – from the web and a further 70 billion tokens from book catalogues. Its impressive results catapulted AI into the mainstream.

Cara Whitehill of Unlock Advisors, an advisory firm supporting early-stage startups and founding teams with seed capital in the travel sector, among others, says, “ChatGPT came out and really democratised access and engagement to that type of technology.”

Scott Wylie, CTO of TMC tech provider TripStax, says ChatGPT has created a wave of interest but also sounded a note of caution. “Business analytics needs to be contextualised along with appreciation that much of the data being modeled will be company sensitive and thus not suitable for use in conjunction with such LLMs.”

Keesup Choe, CEO of PredictX agrees that privacy is a critical concern as well as responsibility for conclusions and recommendations drawn from the data.

“The question is who should be storing that data?” he asks. “There's also another hidden perspective. Most of these chat agents have an indemnification clause that says, in essence, if you copy and paste written text into the chat window and use it for some purpose, you're indemnifying them just from clicking the terms of service.”

AI models binge data – and more is coming

The upshot of that warning is that travel managers may not want to flip an AI switch on their own; privacy – and what some consider the outright “piracy” of open AI learning models that learn from every byte they take – is an especially compelling argument to go with a third-party data consolidator when considering AI solutions. The other compelling argument is that the data feasts required to power good AI tools are enormous, and they are getting even bigger for travel management, thanks to recent distribution fragmentation and increased instances of off-channel bookings.

“At the very minimum you should be integrating the TMC, card and expense data,” says Choe. “The thing that we've been seeing since the pandemic is quite a big increase in in leakage of people booking outside the programme for all sorts of different reasons, for example people who joined during the pandemic who have never travelled before. They may be booking with their personal card to get points and their personal gold Amex or whatever.”

Choe is increasingly seeing direct data connections to both intermediaries like Booking.com and Hotels.com as well as suppliers like airlines and hotels as well as Airbnb and Uber.

Travel companies have also traditionally been reluctant to give up their data, preferring to keep it in their own silos so they can control the customer, but Mat Orrego, CEO and founder of Cornerstone Information Systems says that is going to change.

“Suppliers are going to have to release their data. I think they're going to evolve to have to interact on a broader basis in order to distribute their product and deeply integrate into other commercial platforms,” he says.

TripStax’s Wylie says, “Because of the undeniable benefits of AI, if travel agencies do not change their approach, then corporates will seek their own AI-led solutions as they want better insight for decision and purchasing making processes.”

Jean Belanger, who launched AI platform provider Cerebri AI in 2016, agrees. He said control is at the heart of it. “Vendors like to control the customer. They don't like people getting in between them and the customer, especially big customers.”

Travel data consolidators, like the ones run by Belanger, Choe and Orrego, are more than eager to get in on the AI-action.

PredictX’s AI-powered data platform The Story predates the current AI hype cycle – it was granted a patent in 2019 for its use of recurring neural networks and transformers. The Story automatically generates a narrative with graphics and text after ingesting data from a variety of sources.

The Story can generate output in one of two ways. “One looks like an annual report,” says Choe. “We chose that as a design paradigm as every single big company spends tens of millions in designing their annual report to sell themselves to their investors.”

There is evidence that this form of presentation is becoming less popular and a new variant of The Story is a generative AI platform that can generate ad hoc reports more like a PowerPoint slide deck where users can control how the data is presented.

“The Story in its original form was useful to start to educate the market about what was possible [with AI],” says Choe. “Now we're at the point that I think users are more comfortable and are saying ‘OK, let me control the model. Let me have it generate PowerPoints that I want to generate and let me control the distribution all by myself.”

“Now we're at the point that I think users are more comfortable and are saying ‘OK, let me control the model. Let me have it generate PowerPoints that I want to generate and let me control the distribution all by myself.”

One buyer at BTN’s Corporate Travel 100 Summit in early October said that having “better” data through AI tools like The Story sometimes backfires when it is shared with senior management.

“The only thing they see are the numbers and then they freak out,” the buyer said. “We actually paused the reports because we're going back to the drawing table to figure out how we can tell the story better and give that context and not have everyone hit the red panic button right away.”

Another buyer said that for some business units, The Story provided “too much data” with five-page-long PDFs and some preferred to see a simple infographic instead.

Indeed, simplifying and presenting travel data has been a maddening challenge and getting it exactly right has proven elusive.

When Belanger approached the enterprise data market in 2016 as an entrepreneur, he was looking for something specific: “We had to make a decision what vertical we were going to focus on, so we looked around and asked, ‘Who has the worst data in enterprise?’ We picked travel.”

The goal of Cerebri AI’s AIQ platform is to be “the one source of truth” for travel management. That’s not an original objective, to be sure, but perhaps AI helps the industry get closer to an original solution.

Belanger says the problem is that data is increasingly fragmented and usually sits outside the control of the corporate. He points out that the growth of OTAs such as Booking.com means that oversight of travel spend has been lessened and that AI can help regain that control.

He says, “We take everybody's data and then we pasteurise it, homogenise it, reconcile it, consolidate it – all the things you have to do to make sure you have really good data.”

Belanger says Cerebri AI is working on a Gen AI interface which is likely to drop in Q1 2025 that will enable users to query the data with natural language.

Cornerstone says that such platforms are required because a company typically has more than 400 data sources it needs to consolidate and that the average monthly growth rate for an organisation’s dataset is 63 per cent.

The company is trying to sell the AI dream to travel managers with its AI-powered data platform.

Cornerstone’s Orrego says, “[Travel managers] tend to be looking for anomalies – the differences between the intent of what a travel policy was supposed to yield and what actually happened.” Other use cases for the Cornerstone’s AI platform are contract management and fraud detection.

AI reaches beyond reporting & analysis

Most of the major travel management companies are looking at introducing AI into their workflows if they have not done so already. And the objective, with many of these efforts, is to condense what we think about as “workflow” into automated, AI-informed action.

Amex GBT, which acquired AI startup 30Seconds to Fly in 2020, announced it was setting up an AI innovation team under SVP technology strategy and engineering Michael Esquibel in February.

Esquibel’s VP Marilyn Markham said at the time that AI would “bridge a path to automation previously limited by the language-data barrier” and that it would “unlock efficiencies and inspire product innovation.”

In May, BCD’s Meetings & Events divisions launched its GenAI tool Assist which was “developed to enhance productivity, streamline processes and provide valuable assistance in day-to-day tasks.” It said it was the “first phase of their launch of a robust suite of technology advances being released in the coming months”.

TripStax’s Wylie sees AI being increasingly used in corporate travel. He says fully automated, real-time reporting that continuously learns and improves over time, augmented decision making, highly personalised, adaptive dashboards and self-generating reports, all powered by AI and data, are logical developments.

Unlock Advisors’ Cara Whitehill says the rise of AI is also likely to drive a move towards digital identities and universal profiles that have been talked about for years but may now be ready to come to fruition. Such profiles will include everything about an individual traveller’s behaviour, what their loyalty affiliations are, where and what they like to eat, and more.

“We're moving to a place where the individual is going to own and control all of their data and their persona and how they engage with brands and service providers,” she says. With the right data and analysis, AI might eventually take the place of travel policy which Whitehill calls a “blunt force instrument.”

“We're in this era where I don't think we need a travel policy per se, because I think every trip is contextual. Now we have the tools with AI that can understand the context of why this business traveller is taking a particular trip,” she says.

Some are already looking at this. Cornerstone’s Orrego says, “We have run a pilot with one company where we were able to create a more flexible policy using a chatbot and an approval process within that.”

The yin and yang of our AI future

There’s both opportunity and uncertainty ahead for corporate travel managers as they boldly approach an inevitable AI-enhanced future.

At the Corporate Travel 100 Summit, one buyer from a pharma company said they were trying to bring travel and expense data together to create a policy bot, but the lack of reproducibility and AI’s “black box” nature was a worry.

“I'm terrified because a traveller is going to get an un-reproducible policy answer and now they're going to go and they're going to book a $10,000 business class ticket,” they said. “I have no way of going back and understanding the conditions that recreated that policy answer. That's our canonical source of truth to the traveller because we've stamped it as ‘this is how you do policy questions.’ I'm excited about it, but I feel like unless you've trained it really, really well and you've got the control over it, I'm nervous about doing that.”

In a sign of where AI and corporate travel are going, PredictX’s Choe says he has recently started teaching AI prompt engineering to travel managers so they can configure the tools and generate outputs that are important to them, for example you could configure an agent to analyse contract performance with the right data and the right prompt.

Cerebri AI’s Jean Belanger says the message for travel managers who do not adapt to what is being called the Zettabyte Era is clear. “You're not going to lose your job to AI,” he says. “You're going to lose your job to someone using AI.”