AI transformation in travel: Sejal Amin of Priceline

Rapid evolution of artificial intelligence (AI) is already changing the travel industry and how companies operate both externally and internally.

Over the coming weeks, PhocusWire is checking in with executives about where AI is making the biggest difference in their businesses.

Priceline initially launched its AI-powered chatbot, Penny, in 2023, later rolling out voice capabilities via OpenAI and an interactive AI map. This past June, Priceline also enhanced Penny with Anthropic's Claude. 

Sejal Amin, CTO of Priceline, sat down with PhocusWire to discuss Priceline's AI strategy, how her Penny and her role have evolved, change management and more.

Editor’s note: This interview has been edited for style, clarity and length.

What's your philosophy when selecting vendors as you implement AI?

We've got a number of partnerships relating to Penny. It isn't necessarily about which model wins, but more about what capabilities each model has and what it allows us to build.

We’re using a variety of models where they add value, but the differentiated layer is ours. 

The thing we bring to the table is our proprietary stack, our real-time inventory, our deals technology and everything we know about the customer, which we've been accumulating for over 20 years. 

Partnering with different models gives us flexibility as the model landscape is inevitably changing pretty much every day.

Is it more of a stacking philosophy for you?

It is a little bit of stacking. It's a little bit of bringing the best capabilities that each model has to the table, and taking advantage of those things—it's a traditional build versus buy.

All of the capabilities that the models have, we're not going to build, but we're going to integrate and we're going to take advantage of what everybody has to offer so that we can bring essentially the best the best product to market. It's helped us build Penny's experience. 

How does the addition of Claude impact Priceline's overall AI strategy? Is this specific to Penny, or are there other implications?

We're using Claude not only with Penny, but we're also introducing it to the workforce, like many companies have. And again, it's about optionality and giving not only our end users but enhancing our employee experience in meaningful ways, and giving them access to all the various models and the capabilities those models have to improve their workday, to improve their workflow.

How has your role changed since you started, specifically with the rise of AI?

In my role as CTO here at Priceline, I own product engineering, I own infrastructure, I own data and technology operations. I own the whole stack that our engineers build on all the way to what the customer touches. That's a really important aspect. 

In addition to that, a big part of how I generally see my job is not just the tech, but it's the people. The people manage the tech, and so it's really important to understand how they're organized, so that we are set up for acceleration.

I came in with a mandate to drive AI across our business, whether it be in our products that are serving our customers or for our company that engages our workforce.

What's been interesting for me personally is now deploying tools and understanding that it's not just about tools, but it's about changing how the organization works. 

That's the biggest change: knowing that a lot of time is being spent on organizational architecture and human architecture.

Change management is a huge part of this. How are you handling the education and training side at Priceline?

Education and training is incredibly important. 

The way that we've approached it is a little bit of what I'm going to call it top down and bottoms up. We've taken the approach of deploying all the tools out to people and seeing what people can do with them.

You get pockets of excellence, you get pockets of excitement, you get people who are super engaged and able to accelerate their workflow and you see them do amazing things. It also gives you a sense of what parts of the organization need additional support. 

We did a summer school concept. Every week we had a different training to teach people about AI. It taught people how to apply the tools, and we did training on all of the different LLMs. We taught people how to build agents, and we're observing what takes flight. We're going to watch for the next quarter and revisit and see what's missing, what more do we need.

I'm really hoping with this type of approach, that people start to catch on, they start to see what their peers are doing. We do showcases once a week where teams get to talk about what they built, why they did it, what they learned. Everyone is kind of meeting in the middle.

What level of buy-in are you seeing?

A little bit of a middle of the road, but all in all, I'd say we have 80% to 90% of the organization very much engaged in deploying the technology, whether it be for the customer or whether it be for an internal user. I like to look at it in two parts: AI is affecting our customer experience, and it's affecting our employee experience, and we're investing in both.

What are the biggest bottlenecks you're finding to AI adoption?

People have done great things for their team, or they've done great things for a process or sub-process. The big barrier is stepping back and looking at the entire workflow and engaging people in that and getting them to look at the end-to-end across barriers, across teams, across functions.

Where does competitive difference come in if everyone has the same technology? What’s your moat?

I think models are going to continue to commoditize, and everyone will have access to the same thing. For us, I think our inventory relationships, our commercial relationships, pricing and deal technology, truly understanding our customer, our product design. Those things are a lot harder to replicate. 

How are you handling the risk of error with AI or data corruption? 

With hallucinations, we have a pretty robust evaluation structure we've put in place. We're using various third-party technologies to evaluate the response to LLMs that are provided back to the user for accuracy, for clarity, and over time that gets better and better. 

But we are investing heavily in evaluation technology that continues to check us and continues to check our responses back to customers.

How are you allocating funding to AI? Is investment increasing or staying flat?

The investment is absolutely increasing. You can see in the market we're doubling down on Penny, and we're investing in those capabilities. Our internal investment is increasing in the tools that are being made available generally to people. Our investment is also increasing in our developer experience.

Because all the things I just talked about—our ability to build and accelerating our ability to build—we talked about it in the context of people. But in order for people to do their jobs, they need access to tools, and so our spend in all of those areas has to increase so that we are able to accelerate software delivery.

Where is that funding coming from? Is it being reallocated from other areas? Is it new funding?

Little bit of both. It is coming from existing software spend where we may not need certain tools anymore. We're able to retire those tools, or we're able to consolidate. 

Some of the funding is incremental. It really depends on what the need is.

Do you see employment changing or jobs being at risk due to AI?

I see changes in the way we work, but I think that creates opportunities rather than jobs being lost. I actually think it creates career acceleration. It creates learning. It creates opportunities for people to move into new roles and areas maybe they haven't navigated in before. 

I think people should be, especially in product and technology, excited about change.

Stay tuned for additional interviews with chief AI officers, e-commerce execs and CTOs as part of PhocusWire's ongoing AI Transformation in Travel series. Catch up on previous conversations with Ante Dagelić of Travelier, Velit Dundar of Radisson, Javier Cabrerizo of HBX Group and Piero Sierra of Skyscanner.