Over the past two plus years, Hello Hipmunk has helped thousands of travelers plan trips to all corners of the globe. One of the first-ever travel bots, Hello Hipmunk is also one of the most successful—fulfilling flight and hotel bookings for our travelers with completely automated conversational interactions. Along the way, we’ve learned tons about our travelers’ expectations and interactions with a travel assistant. Those lessons have helped us to build the newest travel bot on the block: Concur Travel bot on Slack.
Launching today, the Concur Travel bot—available in public beta as @Concur_Travel_Beta on Slack—is a new way for business travelers to discuss, plan, and book flights with their colleagues. Our hope is that the bot will save time and simplify travel planning for business trips.
So, how did we get here?
Lessons from an OG Travel Bot
To create a bot specifically for business travelers, we teamed up with SAP Concur and looked at the lessons we’ve learned in the travel bot space. And while two plus years of chatting and travel planning has revealed plenty of insights, these three lessons learned rose to the top:
1. Bots belong in social situations—that are already happening
Whether they’re gathered at the water cooler or sending instant messages, humans want to interact with one another. This, of course, applies to travel planning, too.
With Hello Hipmunk, we did not initially give travelers the ability to share information that the bot found for them. However, when we added the ability to share search results within the conversations they were already having with their friends, they loved it. Turns out, this is also a powerful way to introduce the bot’s helpfulness to more people.
We also found that for many people messaging means talking to friends and family. It’s not natural for people to seek out a bot to chat with directly. They are much more likely to find, and use, a bot if it’s available within the conversations they were already having.
Coupling these learnings with the fact that 40% of flights booked through SAP Concur include at least two travelers from the same company, and we decided to take the sharing and collaboration piece a step further when building our latest bot. With the Concur Travel bot on Slack, business travelers can share search results and travel plans with their colleagues, as well as invite the bot to join in a chat via a Slack shortcut. This helps colleagues search for flights together, in real-time. We think this will be one of features business travelers will enjoy the most.
2. A good bot automates, a great bot anticipates
Who wouldn’t want a bot to take care of a task they dread doing? We experimented with automating the task of finding flight deals for Hello Hipmunk users and found that retention rates increased. This process also taught us that useful bots should automate and anticipate a traveler’s needs.
However, to anticipate a traveler’s needs we need more information—like their travel preferences or historical travel behaviors—to pull from. The good news is that business trips are typically much more predictable than vacations. If someone is traveling for business it’s normally because he or she needs to visit a client, attend a conference, etc. Many business travelers also have specific guidelines (like a travel policy) that needs to be followed. All of this is information that bots can learn from and use to automate tasks and anticipate a traveler’s needs, like booking that monthly trip to Chicago, or making sure that options for the hotel chain someone typical stays at appears in their search results.
3. A chatty bot isn’t always the best bot
With Hello Hipmunk’s natural language processing (NLP) engine, we’ve seen firsthand how travelers actually converse. They can ask Hello Hipmunk anything. We get jokes, status updates, relationship proposals, and more. While we try to direct people back to travel planning, giving them the ability to start chatting with a bot changes the user’s expectations of what the bot can do, and it becomes a race to understand everything that a user can ask.
We’ve also found that when searching for travel, it’s not natural for people to give us all the information we need at once. It’s rare that someone tells us “Give me a flight from SFO to JFK from 8/4 to 8/6 leaving around 8am on United.” Most people chat in short bursts, like this:
- User: “Flight to NYC?”
- Bot: “Where are you flying from?”
- User: “SFO”
- Bot: “When do you want to leave?”
- User: “Aug 4 to 6 – morning”
In a back and forth conversation, the bot has to interpret from implied intent. When it doesn’t know for sure, the bot asks the user for more information. This frustrates the user because it increases the time it takes to complete a task.
The lesson here? People expect bots to be accurate and to understand what they really want—even when it’s not what they’re saying. To help with this, we augmented the Hello Hipmunk bot with visual interfaces, adding buttons, images, and menus. This gave travelers a faster and more natural way to share details and gave the bot what it really needed—specific information it could act on. In fact, when we added a “share location” button so travelers didn’t have to type an answer to “Which airport are you flying from?”, we saw our error rates decrease.
This is why we’re doubling down on visual UI elements in the Concur Travel bot for Slack. Since we’re designing the bot to complete tasks, first and foremost, we’re having travelers start a conversation by completing a search form first. We want to make it dead simple for business travelers to find a flight quickly, and sometimes it’s just faster to click, enter a slash command or fill in a form. With the Concur Travel bot for Slack, speed and accuracy are our top priorities, because we know booking travel quickly and accurately is make-or-break for our business travelers.
So, what does the future hold for travel bots? We believe bots will become a bigger part of travel planning as they get smarter and understand more user preferences. And while the underlying technology continues to improve for bots, we feel the most important work is to focus on completing tasks for people and enabling interactions that are as natural as they are delightful. We’re excited at the prospect of what our cutting-edge travel bots will do to not only modernize the way we do business, but also shape the way we travel.