Innovator vision: The Sky's the Limit
BTN Group’s 2024 Innovate victor is SkyLink founder and CEO Atyab Bhatti. In conversation with BTN executive editor Michael B. Baker, Bhatti shares his vision for travel management
AI-powered corporate travel assistant SkyLink, chosen by the judging panel as the victor at the BTN Group's 2024 Innovation Faceoff in September, was born in part from cofounder and CEO Atyab Bhatti's own experience as a corporate traveller while working at McKinsey and Co. The idea started in an email exchange with friends in 2016 as a consumer product, before the technology was there to back it up, says Bhatti, an engineer by trade.
"You didn't have large language models," he says. "Traditional machine learning was very compute-intensive and very slow, so you couldn't really get to the quality that we have today."
At the same time, Bhatti's time at McKinsey – not only as a traveller but also working with the players across the travel space in consulting – brought him a few epiphanies regarding corporate travel, namely that it also was a massive industry and that it was rebounding quickly. From there, it became clear that SkyLink was something corporate travel had an "obvious need for," he says.
Bhatti spoke recently to BTN executive editor Michael B. Baker about what he's trying to solve with SkyLink, future development plans and what technologies stand to transform the corporate travel industry. An edited transcript follows.
BTN: How did the development of SkyLink come about?
Atyab Bhatti: When you double click into how things are done [in corporate travel], you find that a lot of it is founded on legacy technology. The broader industry adopted aggregators, and a lot of that sits on top of powerful technology provided by some of the incumbent tech players. But a large portion of corporate travel is actually managed through services, and when you double click into these services' workflows, you'll actually start to find that quite a bit of them could have been handled online. On top of that, when these services are being requested, oftentimes it's for non-revenue-generating work for the travel management company.
Around 2021, we took a step back – and I was still a consultant at the time –and I wanted to understand how can we benefit the enterprise, improve the travel experience and align the incentives with the travel management company. Without the travel management companies, corporate travel just doesn't function. You need your BCDs, your GBTs and CWTs. Agents are incredibly valuable and a part of that. What we've noticed is their business model is concentrated on supplier revenue, so their service components oftentimes can get stretched, especially when there are agent shortages. New agents are relatively young, not as fast with the GDS. It's no fault of their own. It's very similar to how we have shortages of COBOL engineers. Similarly with corporate travel, as newer folks come in and more tenured folks retire or reskill into different careers, you're seeing impacts come downstream.
Service levels are coming back and rebounding but oftentimes for workflows that can be handled online. Travellers are frustrated with point-and-click solutions, even with new-age tools that are coming out that are doing a phenomenal job with redefining what it means to have enterprise-grade travel solutions, but at the same time, simply adding a new skin does not fundamentally change the paradigm of booking.
So, the genesis around SkyLink was really around targeting all these problems. How can we make the traveller experience distinctive, improve the economics and operations of the enterprise and do it while supporting the TMC model? That's really what SkyLink is all about. It sits inside your enterprise chat channel where these conversations happen prior to any sort of purchase. It can influence decisions prior to the point of sale by explaining things to users, helping them pick better behaviours, brings a lot of offline transactions online and at the same time supports the supplier revenue for the TMCs, because we help prompt users to actually book hotels when we see them flying a particular route and help them make smarter decisions across the board, so it supports both the enterprise incentives and the TMC incentives. It's about bringing the VIP agent experience not just to your executives but all the way down to that new analyst.
BTN: What sources are you using for content?
Bhatti: We can source content from anywhere. If an enterprise wants a particular content source, the APIs can look different depending on who you're sourcing it from, but they're not that different, so we can source content from any particular supplier that's requested. On top of that, we have pre-built integrations into all the major GDSs and we also have direct connects as well.
BTN: Are you working more with direct corporates or third parties?
Bhatti: We will work with anyone who seems like a mutual fit and is interested in working with us. Typically, that's been direct partnerships with corporates and their respective TMCs. Our core demographic is typically an organisation that has a sophisticated or large travel programme, in the order of millions. It doesn't need to be like $200 million, but a large enough problem that requires dedicated attention. We'll engage directly with enterprises. We've also been engaged with their TMCs as well.
BTN: How does SkyLink handle disruption management?
Bhatti: There are two ways you can think about this: one is reactive, one is proactive. On the reactive side, that's typically when a traveller will let us know that something has gone wrong. We can take an assessment of their trip. An example might be someone needs to fly to Hong Kong next week, but for unfortunate reasons, that trip needs to get moved. You can just tell SkyLink, "Hey, my meeting next week is getting moved." SkyLink can understand all parameters of the trip – your flight, your hotel, whether you've booked other content – and begin to logic through how to adjust that. Maybe that means cancelling your flight and rebooking for the following day when the meeting is actually happening, and then adjusting the hotels in parallel to that. In the event that SkyLink can't do something, we can parlay that over to the agent to handle it. On the proactive part, if an enterprise wants us to pipe in disruption data – a flight's been moved, cancelled, something's getting modified – we can take in that data and map it to users who have booked those particular flights. What that allows us to do is to preemptively let a user know their flight is getting moved and they should address it, or SkyLink can go ahead and modify the flight for you and rebook you later, just like a VIP agent would. It's similar to having United Global Services, but for every user. It depends on the kind of disruption we need to manage and what kind of data we want to bring in to manage the data disruption.
BTN: How are you able to personalise those responses?
Bhatti: There are two ways that SkyLink personalises. Because we know your preferences and your profile, we can start suggesting certain flights with you. If you are Global [Service] with United or the Delta equivalent of that, SkyLink is more likely to serve you up that United or Delta flight, and if that United or Delta flight is not available, we'll serve you up something within the alliance that supports your own incentives, because we know that's what traveller behaviours are. This uses a robust machine-learning algorithm that sees what you're booking along with your preferences along with the enterprise's priorities as well. You might be a Marriott individual [Bonvoy member], but the enterprise has preferred rates with a particular Marriott property, so not only will we serve you up a Marriott more because we know you are a Marriott individual, we will also prioritise the Marriott preferred rate in that location.
This is done through a complex machine-learning algorithm. You can think about it how Instagram or Facebook or any of these major social media companies target you with ads. SkyLink uses a similar logic flow underneath that rapidly understands the different ways you're booking and updates its ML algorithms to not only understand your preferences but the enterprise priorities as well and pushes users to do this.
Another thing, because SkyLink can intelligently help you book, we can figure out if you're not booking a hotel on a particular trip and prompt you to do that. It's deeply complex for us but extremely simple for the user. For the user, it's as simple as ask a question, click a button, get your booking done. For us, there's a lot of engineering happening on the back end to make this to make this such a wonderful experience for the traveller.
The second way SkyLink learns is that as you talk to SkyLink, it begins to have a memory form of your preference, what you're doing and what you like, and it uses that memory function to adjust how it communicates with you. That uniquely allows SkyLink to calibrate itself across two main dimensions. You can think of it as an EQ [emotional quotient] calibration and an IQ calibration where the IQ is coming from the machine learning and the EQ is coming from its memory capabilities.
BTN: What's on your roadmap for future development?
Bhatti: What we're working on is really designed for sophisticated programmes. If it works for those sophisticated programmes, it works for pretty much everyone. The way that we prioritise our roadmap is really about what we think travellers need. Once we figure out what the traveller needs, we work backwards toward what helps the travel manager in this context and what helps the TMC in this context.
Today, you can ask SkyLink to book content for you, you can ask it to modify content for you and you can also ask it non-travel-related things, like helping figure out what to do for a business meeting, what restaurants are out there, what the weather might be like, whether there's a disruption or a policy question. The future isn't going to just be in chat but it will evolve into different modalities. SkyLink will be exploring and leading those kind of frontiers for enterprise-grade shopping.
BTN: What wider benefits do you think the corporate travel industry is going to see as generative AI proliferates?
Bhatti: One of the big ones will be the shift in productivity. It will be similar to what Google did for finding information. There will be use cases where people will build very powerful tools with Gen AI, but if you look at it from the broader population lens, it will largely be an augmentation to the human productively.
In the context of corporate travel, what that can mean is knowledge that is specific to what you are looking for is constantly available. That will be the big adoption curve that happens. We’re also going to see a component of verticalised agents. SkyLink is doing this in the corporate travel space, but building vertical technologies is quite challenging, given the nuance of every industry. We'll see domain-specific tooling come out, and that will have a huge impact on how productivity and labour distribution happens, but it will largely be beneficial.
We've seen time and time again, technology drives more productivity but a better quality of experience in life as well, and Gen AI will have a similar impact as a paradigm of the internet and computing and atomic energy. A lot of that is going to follow through a similar path.
What will be different this time is the rate of growth will be far faster. It's going to be unlike anything we've ever seen. Gen AI hits pretty much every kind of lever that is needed for adoption, and it has infrastructure laid out for it. Unlike the internet, where you had computers but may not have had landlines fully hooked up yet across the world, we had to build that infrastructure, which delays adoption. The same with mobile. You had mobile phones that are very advanced, but not everyone could afford them because the costs were high when they initially launched. As infrastructure went up and costs came down, they became mass-adopted. With Gen AI, everything is already there, and so adoption is going to be exponential.
BTN: How will this change the skill sets required to work in corporate travel?
Bhatti: It's a little tough to give a blanket answer. Travel agents will get significantly better, in the sense they'll have an assistant who can tell them GDS terminal commands if they need to. You'll see their speed and velocity improve, so long as the large TMCs embark on this kind of change, and I think they will.
On the travel management side, things will get a little more efficient in that you won't need a data analyst to look at all your travel data. You will be able to query your data using some specific vertical IS tool that knows how to do all that work on behalf of what a data analyst and travel manager would do. Relative linear tasks will become automated.
Where Gen AI may face a little bit of challenge is deeply complex tasks. Even as inference improves and intelligence and reasoning improve, we'll still run into nuances that only humans can really understand today. A generalised piece of technology might get better at it, but you'd still need a human oversight. In the travel industry, broadly, everyone will get more productive. Jobs will get augmented. Back in the 1800s, someone who was really into maths was running all the calculations in their head. Now we're all PhD-level mathematicians, because we can have our iPhones out and write those numbers down.
BTN: Besides AI, what technology has the potential to transform the corporate travel industry significantly?
Bhatti: One of the big ones is quantum computing. Typically, computers work in binary functions – zeroes and ones. Quantum does a concept called superposition, where you can exist in all possible states until you are observed, and then you enter into that quantifiable state. The 'so what?' of why that matters is that you can calculate every single outcome using quantum computers much faster than you can with traditional computers, because you can run multiple calculations simultaneously. That has a massive implication on our ability to solve mathematical problems and do research. With Gen AI and eventually the advent of quantum, that will be transformative.
There are downstream, sub-innovations that have huge impacts as well. Last year, there was LK-99, the room temperature superconducting, which has a huge impact on quantum, because quantum computers require extremely cold temperatures to bring the qubits into superposition. Those two things as a whole will have a massive impact on society. We're working on that, but it hasn’t happened yet.
Who knows with these smarter Gen AI models, that they'll just run research 24/7 and figure something out. The next big one is really around energy. Today, we use a lot of coal and non-renewable sources, and we are seeing renewables take off, but energy [has] becomes more cost-effective to source from a renewable lens. Solar panels are becoming more efficient.
Once we see the popularisation of fission and fusion reactors and nuclear power becomes more popular, that will have a massive impact on corporate travel. If you can miniaturise a reactor, and there are safety implications of putting that inside a plane or something like that, energy efficiency will have a massive impact. You could effectively modernise every bit of technology because you had renewable technology. If you could replace non-renewable fuels with that, you'd not only make the world a better place by making it more green, but you'd also have a direct impact on the cost of fuel, which can impact corporate travel.
BTN: Consumer travel technology has typically drawn more attention from investors than corporate travel technology. Are you seeing any shift there?
Bhatti: From my perspective it hasn't really changed. Corporate travel is not latent because of investor interest or lack of investor interest. It's rather latent to the counterparties involved in actually executing what corporate travel is. The incentive structures are different for every party, and that pulls people in different directions. Because of that, any sophisticated investor would recognise the challenges that presents, and the only way to make it an attractive space for large-scale investors – venture investors specifically – is the support of largescale enterprises who are adopting new technology, and that's what really moves the industry forward. We're already seeing that to an extent, as some of the large-scale players are exploring new technology. It's baby steps, but eventually baby steps add up and then you're Usain Bolt running the Olympic marathon.
Built with Shorthand