Working inside a large organization, it can be smart to launch new technologies and initiatives in phases, similar to planes on the tarmac waiting to take off.
That’s what the Berlin Brandenburg Airport (BER) passenger experience team did this year when it launched its AI-based customer service agent. This around-the-clock, voice-based AI phone service is available in English, German, Polish, and Spanish. And it has not only attained an 85 percent approval rating among BER customers, and reduced customer wait time from around two minutes to zero, it also won the Most Innovative Airport Initiative Award at the Future Travel Experience industry fair in Dublin, Ireland, in June.
Building on that success, the airport’s passenger experience team is working on phase two of the project — including a chatbot and QR codes — just months after its initial launch in February. Christian Draeger, Senior Vice President of Passenger Experience at BER, recently spoke with InnoLead about this innovative approach to customer service at the third-largest airport in Germany.
My Role
Christian Draeger: My role is to look after our customers, and our team is the main advocate at the airport to make customers feel happy and give them a great experience. And how are we doing that? We have [several] departments that are part of the whole passenger experience as they navigate through our airport.
One team looks after land-side topics like, “How do I get to the airport?” We also have a team that oversees our terminal services, which encompasses the entire customer experience as they go through the airport.
Our large security control department is responsible for ensuring that the security controls customers undergo are conducted in an efficient and friendly manner. And we also have a team looking after our technology topics and processes… where we need to go in the future to be relevant and provide the right services to our customers.
The icing on the cake is our premium services department, which operates two business-class lounges at the airport that serve about 400,000 passengers a year. This department also offers an ultra-exclusive VIP facility for customers seeking privacy and the best in food, entertainment, and services.
The Project
CD: We weren’t entirely satisfied with our call center’s scalability, hours of operation, and the number of languages it offered. It was rather limited. We also felt that financially, it wasn’t where it should be, and that financial part also played a role in our exploring what kind of alternatives were out there.
When we considered all of that, we said, “Let’s try to find an alternative technology solution that provides better service to our customers in terms of information,” since one of our key responsibilities is ensuring that customers at every stage of their journey have the right level of information, and also one that allows us to save money. That was basically the preconditions, and the reasons why we went into this.
Two things really worked well for us. First, we defined our problem statement from the outset and stayed focused on resolving it. Our problem statement was initially to reduce cost, increase revenue, and provide customers with better information. We didn’t try to add 15 different other aspects to it at the same time.
The second thing that helped us was breaking things down into different phases. We started with our initial problem statement as a first step. But if there was a great idea out there that just didn’t fit our initial step, we set it aside and said, “Let’s not forget about it, because there might be a place for it in phase two or three of our project.”
How It Got Green-Lit
CD: If you can demonstrate you can create value and a robust business case, it would be a great help. But if you could only argue that it would increase the customer experience, it would have been a very different story. That’s because customer experience is sometimes hard to quantify. You need to quantify it with facts from day one. We were able to demonstrate that we were actually creating value and saving money, which was one aspect of how we got it approved.
Having a sound business case was also really helpful. Our project had a good and relevant impact on our customers by providing them with timely information, while maintaining a very low-risk profile.
…It didn’t require a huge and heavy investment. We needed to make choices about the type of technology we wanted and its capabilities, and the one we selected did not carry risks and liabilities that were extremely complicated.
Additionally, it didn’t require a huge and heavy investment. We needed to make choices about the type of technology we wanted and its capabilities, and the one we selected did not carry risks and liabilities that were extremely complicated.
For example, in our initial phase, we could have added personalization to provide an even better customer experience, but it would have required us to gather more personal information, and there are complexities and risks associated with data protection issues, AI rules, and obviously, the risk profile would have been slightly higher.
The third reason we were able to convince our board and chief executives to approve the project was that we not only explained the benefits of today and tomorrow, but also outlined a roadmap for the next three to five years. They could see where this was going, and by investing in this project now, we would get ourselves future-ready for the significant, game-changing personalization that agentic AI will bring to the way we travel and live. Having a long-term plan helped us secure project approval.
The Timeframe
CD: It took three to four months for the idea phase alone, followed by another six weeks from the “yes” decision to launching the project.
The initial steps involved understanding what kind of technology was available, what it could do, how it worked, and whether it fit into any of our problem statements. We were also exploring potential partners, because we knew from day one that it was essential to have partners, given that we’re an airport provider and recognized that we wouldn’t be able to do this alone. Although we were initially unsure whether to proceed, we eventually made a decision once we were convinced we had learned enough about the technology and the playing field of potential partners.
…From day one, we were able to identify who within our organization could play a crucial role in building the preparatory work that needed to be done, such as the knowledge base. Our partners, Parloa, brought large language models to the table and the technical solution, and KINOVA provided the technical implementation collaboration. But we needed to do our own work to put everything we knew about the airport and its customers into the knowledge base. We had multiple departments and multiple people working on building this knowledge base, which now serves as a reference base for our AI to train on applying algorithms.
After it launched, we continued to use our previous call center provider for another six weeks as a safety net. They were ready to jump in if any large problems or bottlenecks occurred, which didn’t happen. So, after six weeks, we concluded our relationship with the call center provider, and the project has been operating on its own.
The Next Phase
CD: We are now ready to jump into phase two of our project, which will introduce chat functionality. Currently, our voice service is primarily designed to provide you with information to help you prepare for a trip, typically a few days before the actual trip, when you’re still at home. But with the chat functionality we’re developing, customers will be able to access it at our airport, with much of our focus on the day of travel.
The chatbot has two channels that are always in service. One of them is our web page, and the other one is our Berlin airport app. But we know that many customers don’t download the app, especially if they’re visitors to Berlin and only visit once a year or perhaps only once in a lifetime. So, what we’ll be providing in the future are QR codes at 40 to 60 different relevant points in the airport, as well as in the surrounding environment, such as adjoining hotels or public transportation trains. With the QR codes, customers can immediately access the latest gen AI agent to chat about flight delays, rebates at the airport retail stores, or access our purchase platform.
The Three Biggest Challenges
CD: One of the initial challenges was really understanding the tech and what distinguishes one offer or provider from another, and where some of this technology is heading. Even with a company of our size, with around 2,000 to 2,500 employees, we don’t necessarily have a deep understanding of all aspects of tech.
The second challenge was particularly important to resolve. We needed to find the right partner. There were a lot of different choices, and we had many attributes we were looking for. Although we didn’t have an opinion on whether our partner needed to be from the U.S. or another country, funnily enough, the technology partner we found, Parloa, was based in Berlin, 20 minutes away from our office. They are in Germany and Europe and are one of the leading companies when it comes to agentic-AI platforms.
The third challenge was not to lose momentum once the initial solution was deployed. When we did the voice part for the first phase, we had a very high level of support and excitement with everyone in the organization. Now, as we enter the second phase, it’s essential for us to maintain this level of excitement and not assume that the first phase is complete and the job is done. No, it’s not done. There’s so much more potential for our customers to experience.
The Smartest Thing We Did to Set It Up for a Successful Launch
CD: The partner selection was really the smartest move, and a counterpart to that was our internal stakeholders. We excelled at achieving full alignment among all departments. To build the knowledge base with relevant information, you basically need to touch every single department in the organization. We were able to maintain a positive dialogue and achieve good alignment among all internal stakeholders from day one.
Metrics We’re Tracking
CD: During the six-week preparatory and launch phase, there were a few things that we always kept in our vision.
The first thing was around savings, and that’s easy to track. The second thing was speed. From day one, we really stayed very close to our timelines. During our daily stand-ups, we would identify any bottlenecks that were derailing our timeframe. That focus on speed really gave us a clear understanding of how we developed the project.
We also knew that if the knowledge base wasn’t at a certain level, we wouldn’t be able to launch it, and we needed to ensure that the knowledge base’s richness and maturity were at the right level to proceed.
From day one, we kept our focus on ensuring intuitiveness when customers interacted with our AI agent and how those conversations were conducted.
The final metric we measured may sound pretty mundane, but I think it shouldn’t be underestimated. From day one, we kept our focus on ensuring intuitiveness when customers interacted with our AI agent and how those conversations were conducted. That’s because we felt this would be something that could hinder people’s use of the service. If the dialog between a customer and the AI wasn’t at the right level of intuitiveness, we observed a significantly higher drop-off rate of customers, even during our testing cycles.
[Now that the product has launched,] we have two kinds of metrics. One is performance, or the output of the AI agent’s work. The other metric focuses on the solution itself, such as how it evolves and grows based on the information it’s exposed to in the background.
The first metric is really about hard figures like the number of calls we now receive, which has increased by a factor of four since using the AI voice service. Our average wait time to get help has also dropped to basically zero from about two minutes prior to the switch away from the traditional call center service. Our customer satisfaction rate, meanwhile, is 85 percent, but it’s not really where we want to be. We want to be above 90 percent. However, we’re heading in the right direction, and as we train the agent, we’re seeing its capabilities increase more and more.
The Future
CD: Although AI can currently provide information, there is a whole layer of decision-making and action-taking that is still down the road. However, once this starts happening, AI will provide customers with a much more personalized experience.
Customers may want to have their own individual AI agents. This is not necessarily an AI travel agent. It is a personalized AI agent that can provide its owner with different levels of support for different areas of their life. The AI agent might handle all of your financial, health, and travel matters.
For example, my personalized AI agent will know everything about me. It will know my schedule, my favorite way to get to the airport, and whether I’m prone to shopping at the airport. With this knowledge, this agent can start acting on my behalf in a much, much larger way. As an airport provider, we need to be ready to talk to these agents in the future and provide them with our content and solutions. Humans will no longer be the end customer. We need to really look at the future as humans, plus AI agent. Whether all of this will come to fruition, I cannot guarantee, but that is one of the thoughts that we’re trying to follow.
You must be logged in to post a comment.