Aurinkomatkat x Evolver: AI is revolutionizing Travel Consulting

Evolver and Aurinkomatkat embarked on an ambitious joint project that resulted in an AI-powered travel assistant.

Aurinkomatkat and the software development company Evolver jointly created an AI-based travel assistant that, as it evolves, could transform customer service across the travel industry. A bold culture of experimentation and technological leadership have already delivered tangible business benefits.

Aurinkomatkat is Finland’s largest tour operator. The first online store was launched more than 20 years ago, and today about 96% of all sales happen through digital channels. Even so, customer service has been under heavy strain, especially with questions about choosing destinations and hotels.

“A large share of our customer service contacts were about this. We needed a way to scale expert travel consulting digitally,” says Ilari Pohjola, Head of eCommerce & Marketing responsible for developing digital services at Aurinkomatkat.

The need for a scalable solution was clear—but implementing one safely, without weakening the customer experience, was not straightforward.

The right moment arrived as generative AI and large language models advanced. Together with long-time partner Evolver, they launched a development initiative to build an intelligent AI assistant to support travel consultation.

From customer need to AI solution

The project began with an internal review to develop self-service solutions. That work revealed that a large share of customer service contacts related to travel advice.

“Customers want help choosing destinations and hotels, and we know that good advice improves sales,” Pohjola sums up.

At the same time, generative AI and large language models began to progress rapidly. Together with Evolver, the idea emerged of an AI assistant that would offer a scalable alternative to human consultation.

“We launched an AI-powered travel assistant that lets customers plan and book holidays by conversing with a smart AI assistant—it’s a far more advanced solution than a traditional chatbot,” Pohjola reflects.

Evolver CEO Arttu Kataja calls the collaboration a true innovation project, grounded in solid research, an understanding of customer needs, and a willingness to take calculated risks:

“Aurinkomatkat’s online service is at the heart of the business. The needs were identified through research: customer service required a scalable solution. At the same time, AI technology and language models advanced quickly, enabling a new approach. The project was started boldly, even though the full scope of the challenge wasn’t clear at first. As AI developed, we were able to build a functional solution architecture.

Once we had a sufficiently capable base model, the right architecture, and a grasp of everything that had to be considered, the work became relatively straightforward. There were moments of despair along the way, but the end result is an excellent, functioning service to be proud of.”

A calculated risk—with rewarding results

The development work required internal quality checks. After several iterations, the AI assistant was first released to a limited group of users and later to all visitors to aurinkomatkat.fi. Going forward, it will handle up to 100,000 conversations per year.

“We did a lot of internal testing. Especially during the build phase, we aimed to validate the quality of the conversations. That took quite a bit of time. Next, the service was rolled out for internal use by customer service. We also ran a closed beta, inviting the developer community to test the service,” Pohjola explains.

“Customers come to vacation dreaming from different angles: some already know the destination and start looking for a hotel, while others seek ideas based on what they could do at the destination. The AI assistant helps find the right destination and hotel based on the customer’s preferences. The recommendation depends a lot on how comprehensively the customer describes what they want from their holiday,” Pohjola reflects.

Starting development took courage, because this is a service that operates directly at a large company’s customer interface.

Over his career, Pohjola has seen passing trends, but the advances in generative AI and language models felt like a bigger shift. Watching from the sidelines wasn’t an option—this was something to be part of from the beginning.

“I’m glad we had the courage to move in this direction. Although a direct-to-consumer AI implementation carries its own risks, we succeeded in reducing them significantly through various measures. It’s a controlled risk, where the potential outweighs the risk.”

Kataja praises Aurinkomatkat’s pioneering mindset: “It was a similar situation when the internet arrived and you were among the first to take commerce online. What if you had waited then? That would probably have had an even higher cost.”

Distinctive features of the development

AI initiatives have their particularities, even if in broad strokes development proceeded like any software project. There was a business need, strong customer understanding, and thorough discovery and concept development behind it.

According to Kataja, developing a deeply integrated solution amounts to a full-scale IT project.

“What’s new, however, is that the early phase involves genuine, experimental innovation. You can’t do that with the same heavy process as production development—but both are needed in this kind of effort,” he emphasizes.

“The team also had to learn a lot of completely new things. Traditional job descriptions went out the window—for example, designers learned how to prompt,” Pohjola adds.

“In the beginning, large language models—generative AI—didn’t enable everything we wanted from a technology standpoint. Still, we trusted the pace of progress, and that’s exactly how it played out—the leaps in the models ultimately made the desired outcome possible,” he adds.

Risk management front and center

The risks associated with adopting AI were considered from the start of development. With the right choices, risks can be managed and quality assured.

“We’ve built a solution that relies on our own content—for example, hotel information comes directly from our content. We also use automated testing that runs through hundreds of test conversations whenever the system is changed. AI is also used to help ensure quality,” Pohjola explains.

He concedes there is no perfect risk control: “One conversation in a thousand may go wrong. We’re transparent with customers: we tell them they’re talking to AI and remind them to double-check the information.”

Regulation and privacy were also central to the design from the very beginning. In this case, the AI service does not process personal data—the conversations are anonymous and the user cannot be identified. Even if more data were involved, the architecture is designed to prevent problems. Under the EU’s AI Act, the solution is also classified as a low-risk case, because it does not process sensitive data.

“Privacy and regulation are always on the table in this kind of development. If the case were categorized as high risk, I wouldn’t recommend starting there. Here, the business need and the requirements aligned well, so regulation didn’t pose any particular challenges,” Kataja says.

The solution delivered tangible business benefits

AI-assisted customer service has eased the workload on customer support.

“We’ve taken a significant amount of pressure off our customer service, as a large part of the consultative selling around trips is now routed into these conversations,” Pohjola says.

The high volume of conversations and strong customer satisfaction are signs of success. It shows up in sales, too.

“The solution improves conversion; on peak days, nearly a third of conversations with the AI lead to a sale,” Pohjola is pleased to note.

Evolver CEO Arttu Kataja emphasizes the importance of design in delivering a great user experience. A small animated line of text invites even a first-time visitor to start typing. Even small things have a big impact, a point with which Pohjola agrees: “A user may never have chatted with AI before. Even small interface decisions make a huge difference.”

The solution aimed for scalability, efficiency, and accessibility. The feedback has been encouraging.

“The goal was round-the-clock support and a source of inspiration for booking a trip. Typically, customers praise how convenient it is to narrow dozens of hotel options down to a few suitable ones,” Pohjola says.

The journey together continues

Yksi projektin keskeisistä opeista on vaiheistuksen merkitys. Kunnianhimoisessa projektissa rajoitetumpi, ns. pienin toimiva tuote (MVP) olisi voitu julkaista aiemmin tuotekehityksen tueksi. Se olisi nopeuttanut oppimista.

“Alkuperäinen tavoitteemme oli erittäin kunnianhimoinen, vaikka teknologia ei ollut vielä täysin valmis. Projekti olisi voitu pilkkoa pienempiin osiin ja julkaista kokonaisuus hieman keskeneräisenä – ei laadun, vaan laajuuden osalta.”, Pohjola summaa.

“Tämä ei ollut hyllyltä valmis pakettiratkaisu – me laitoimme kädet saveen. Nyt meillä on työkalupakki, jonka avulla voidaan rakentaa paljon muutakin”, Pohjola kuvaa.

Nyt rakennettu perusta mahdollistaa jatkokehityksen esimerkiksi matkan aikaiseen palveluun. Pohjolan mukaan mahdollisuuksien kirjo jatkokehitykselle on valtava, kun asiakaspolkua katsoo matkan varaamista laajempana kokonaisuutena. Yhteinen matka jatkuu.

“Kun olemme Evolverin kanssa toteuttaneet tätä ratkaisua pioneeri-hengessä, niin kaikki opittu on meille arvokasta valuuttaa ja synnyttää kilpailuetua.”

One of the project’s key lessons is the importance of phasing. In an ambitious project, a more limited minimum viable product (MVP) could have been released earlier to support product development. That would have accelerated learning.

“Our original goal was extremely ambitious, even though the technology wasn’t fully ready. We could have broken the project into smaller pieces and released the whole as slightly unfinished—not in terms of quality, but scope,” Pohjola sums up.

“This wasn’t an off-the-shelf package—we rolled up our sleeves. Now we have a toolbox we can use to build much more,” Pohjola says.

The foundation now built enables further development, for example for services during the trip itself. According to Pohjola, the range of possibilities is vast when you look at the customer journey as something broader than booking alone. The shared journey continues.

“As we’ve implemented this solution with Evolver in a pioneer spirit, everything we’ve learned is valuable currency for us and creates a competitive edge.”