‍

Akemi opens up by explaining that her business is founded upon the need to respond in real-time with technology solutions to crisis management needs. Another priority of her business is 'flow management', where she works with destinations to help them manage and try to overcome over-tourism. Today, the same feature is used to stimulate in-market spend and contribute to social distance.

‍

The coronavirus outbreak was something unique but that she felt in some ways comfortable with: she explains that for her, being Japanese, crisis management always meant natural disasters. This is the experience of Japan, whether it's tsunamis or earthquakes there are always challenges to overcome, which is where her experience stems from. However, the coronavirus pandemic poses a different problem and requires a different response.

‍

Working with AI, chatbots are there to respond to common questions and in the context of a crisis, common questions can really help respond to people's current problems. Akemi describes how the volume of incoming messages spiked exponentially during the crisis disproportionately to the rise in confirmed cases. This started with "where can I buy a mask?", quickly moving to more travel-related questions such as "If I have a Chinese passport, can I travel to your country?", then evolving to more complex questions, such as "If I transit through country x, and I come from country y, can I visit Japan?". The complexity of queries just continued increase as the pandemic evolved.

‍

Later on, the messages became even more challenging to respond to, with people's real health problems being described in the chat and people seeking support or medical advice. So what is evident with all of these examples, is that providing an immediate response is key. We can also see that not all emergency services can be automated, sometimes you need to speak to a human. Not all information is available in all languages and a chatbot can't have the answer to every question.

‍

Looking at data from September this year, we can see that almost 40% of the queries were related to travel destinations, nationality, country of origin and self-isolation rules. Akemi has seen things evolve over time, with the sentiment of queries improving as the virus becomes more embedded in society and people focus on more practical questions. This gathering of conversational insights allows Bebot to identify trends and feed that back to government and decision-makers.

‍

Looking at the wide range of queries coming through chatbots helps the team at Bebot to craft a better response. For example, telling people simply that 'today, 200 people tested positive for covid-19', gives a very negative message. A lot of the actual questions related to restrictions, masks and rules and a better way to communicate would be to respond to the actualquestions that people have. So a better way to craft a message is to respond to their actual needs at that moment in time. The challenge is that what the travel industry wants to tell travellers is more often than not different to what travellers want them to communicate.

‍

To learn from an example of communicating around a different crisis, we can look back at the typhoon in 2019 when 17,000 passengers found themselves stranded with land transportation cut off, including the main highway to the airport. So rather than leaving passengers stranded and helpless, the team went through all the queries coming in through chat and looked at how they can respond better next time. They saw that the volume of queries varied over time. More than 50% of the queries relating to land transportation was about a particular train company going in a specific direction. Others were related to how the different airlines supported their customers with variations between Japan Airlines and All Nippon Airlines, pointing to how they supported their customers at that moment. Other complaints were related to vending machines not accepting credit cards, whilst the passengers in the terminal came from many different origins and didn't have currency, clearly pointing to a range of problems to address.

‍

This critical intelligence role of a chatbot is something that should not be underestimated alongside the ability to respond with greater relevance and accuracy to common queries.

‍

Akemi also points to the fact that in some cases, a chatbot can respond faster than a human. She gives the example of a chat query typing 'call 911', where a mother and child were in a critical situation with the child not breathing in the room. The mother couldn't communicate with the front desk in English, so she turned to the chatbot which seemed better than having no support. For Bebot, it demonstrated an opportunity to combine the reliability of a chatbot and a human to create a very powerful solution when it is used properly - to raise alerts or support people in a crisis.

‍

If we look at travel and tourism, there are also more positive opportunities where chatbots can play a role in helping shape and improve customer experience. For example, recognising common queries around pool opening times can help the hotel management understand the level of demand and respond to that in order to improve customer satisfaction. Online too, there are learnings we can gather from chatbots, analysing feedback such as "your website isn't easy to navigate", and addressing problems to optimise the user experience.

‍

Lastly, another common query which showed up as an odd enquiry to receive through the chat is people asking Japanese hotel sites if they can book for 10+ rooms. Investigating this issue further, it became apparent that many hotels didn't have the possibility of booking more rooms, yet the demand is there and likely to go elsewhere if not responded to. This kind of intelligence allows brands to apply user feedback to the digital experience and avoid missing huge lost revenue opportunities.

‍

Closing off, Akemi shares the example of Vienna Airport. Currently, over 20% of the queries are of people enquiring whether cafes and restaurants are open. Other common queries relate to rules around the wearing of face masks or providing transfers even if you're not flying, so learning from this helps us to respond to common queries.

‍

If you want to think about how to prioritise your investments in chatbots, think of it like this. Spend 20% of your time and budget on making a bot that works, spend 30% promoting it and spend 50% actively engage users and learning from this and iterating.

‍

Lastly, there are so many queries that we simply can't answer all of them. Focus on the volume, not the long-tail as that's where an automated response can serve best. If you focus on the most common questions, you can do a better job of creating a workflow of automated questions and sending users through a logic-based journey to get all the answers they need.

‍

So whilst a Chatbot may seem very simple, there's a lot of technology, engineers and work sitting behind it. Right now, a chatbot for the travel industry is very vertical and narrow. In the future, expect there to be a wide range of chatbots which are very tailored talking to other chatbots and better serving the needs of users with a greater level of accuracy as they learn and improve.