Case study: The Future of Chatbots: ROI, Customer Loyalty, and Revenues
Thanks to Federico Gobbi from @AIMA for the write up from the AI Expo North America 2017 (29–30th November)
By 2020, 85% of all customer interactions will be handled without a human agent, someone said. But today we are in 2017, almost 2018, and there’re still a lot of doubts when it comes to chatbots.
On Thursday November 30, I served as the Moderator for the panel “Chatbots: The Next Generation of Messaging Apps” at the AI Expo in Santa Clara.
Great opportunity to hear about some relevant case studies in this field from experts such as #eBay Lead Product Manager, #MoneyGram International, Head of Product and Innovation, #TGIFriday Marketing Technology Director and #Foursquare Director of Consumer Products.
The session explored such key topics as:
- The rise of AI and the rise of chatbots
- How to set up chatbots to meet business goals, particularly goals around customer loyalty and increasing revenue
- Best practices for humanizing chatbot responses for optimal consumer interaction
- How to involve the leadership in your chatbots crusade
Give us a little bit of introduction to what you guys do and what is your experience in using chatbots? Also, if you are already using or planning to deploy a chatbot?
We are leading an AI-powered shopping assistant. Most recently we had a chatbot on messanger and we very recently launched a eBay assistant on google home. And before that I was working with Stubhub building chatbots on Messenger, Skype, and Cortana.
In addition of running our mobile apps we launched a year ago a chatbot that converse with you via SMS to find drink recommendations and also, it does contextually powered personalized recommendations
We deployed our chatbot about an year ago. His name is Fry. We are using Fry on twitter and facebook, as well as the Alexa skill.
We are engaging in a very aggressive digital tranformation. We launched a bot that we co-designed with Facebook on messanger. Also, using AI to anticipate the behaviour of the users as they are going through a very emotional journey sending money.
The following questions were asked to the panelists:
Challenges of building chatbots for the first time? and let me play the Devil Advocate role, can you share stories about how things have gone well and bad?
What we realized was a good strategy when failing was providing a smart message to still deliver value. For example when a user asked something that our chatbot is not able to understand due to a lack of data — e.g. customer looking for tacos adding also another part in the message that we were not able to process — we manage to provide a suggestion based on the word we were actually able to grab by making it saying “I think you said (tacos) why you don’t look at this (a solution, maybe not the customized one). And we realized that the customer was actually happy and engaged versus saying “hey, I don’t know what you just said.” So failing smartly was our first challenge.
How do you inject personality into your chatbot? How differently you use your chatbot platform from the regular versions?
It does really start by how do we want this chatbot to sound like. And we want to sound like a closed friend that you know really well and it recommends place for new ideas. It’s actually all written by me and my colleague thinking about how we naturally interact with people. We did research on it and one of the funny ones was texting to the person “hey, what are you up to? Do you wanna check out this restaurant in the mission, thought was really great and I know you love indian”. And I remember one of our research said that when she got the message she thought it was someone she was dating and asking to go to the restaurant. So we realized this was a little too closed, too personal. So we detached it a little bit but we keep that personality adding emojis, abbreviations, to try to make the all idea like it is a human that you are texting with. I think that is one of the differentiator in chatbots. The way you engage with people in a human level. You really need to think about your chatbot in a very different perspective and understand how is really perceived by your customers.
What kind of issues or challenges chatbots can solve?
We went through the same learning curve that Marissa just highlighted. So what we have done is also using a funny image to eliminate the stress in not providing the answer. And then, we moved forward to suggest something by looking at the customer profile and suggest based on the previous experience. Try to divert a very frustrating experience to suggest what the user would like to do. Use data and previous behaviours.
How important is for your company to provide an answer to user questions real-time?
- It’s very important. When somebody sends money they want to instantly know if the received got the money. And if they won’t know it they will call the customer service. And if they are going to call me we are ruining our margins — considering the cost of a call — and lose the profitability. So the ability to have mecanism where they actually know. Typically what we have done is when the sender process the money we send back a message saying “Ok, done. Money has been picked up.”. Another thing we realized is that a lot of time people don’t event thank the sender, due to the different salary situation. So what we do is sending a thank you message, which it was impossible to do through an app before. In conclusion what we have seen is that people don’t go back to the app or the traditional methods. They want to chat with our chatbot.
- There’s still a challenge, which is something that the bot cannot do. So far has been hard for us to transition from the conversation with the bot to the real person, such as in a situation where the bot can’t answer a question because it is impacting highly regulated rules that a person has to manage.
What are the metrics you are looking using chatbots? How are you planning to scale it?
When we first started we implemented our bot to decrease interactions for our social team. And then we moved forward supporting the order system and we were able to get more insights in terms of customer service as well as payments and transactions.
What are the chatbots capabilities you will like to use?
What we are really doing today is customer service, and online ordering. What we want to do in the next future is improving the customer acquisition. So far, we have seen very good results and we look forward to understand how we are able to make it smarter, especially making sure we’ll be able to recommend something that can be actually relevant for our customers (we want to avoid to recommend chicken to someone that never ate chicken before.
How to convince the leadership to deploy a chatbot and allocate a budget?
- It really comes down with the 3 key major points you should provide to the leadership. I heard some great quotes, like Ai will be the new UI. And if we break it down there is a lot of truth to that. Applying AI the new user interface will be pretty natural, more human, either if it’s going to be through a chat or voice conversation or image and so on. So what this mean is that if you believe this is happening you have to make investments as soon as possible. We definitely believe it’s a window opportunity and we making this case pretty clear.
- We don’t have just to look at the long term impact. There’s a lot clear short term impacts that can happen, such as reducing cost, customer service, acquiring new users and there’s a lot of important KPIs that can be shown in a few months especially talking chatbots wise.
- The last part is actually the cost is not that bad.There’s a lot of off the shelf stuff you can go after to get the ball rolling and get to a proof of concept, prove that what you have done works and then you can present it to the leadership and make a larger investment as well as build an entire AI stack.
Youri: moneygram network has 350K+ agents worldwide that can pick up cash. And we realized that actually people challenge the chatbot trying to find places where we didn’t have locations. And so, what we did was starting to put images or emojis, e.g. picture of the desert that says middle of nowhere. And after this people were putting actually the real location. People that send money are really careful on what to do, so the chatbot conversation is actually very important.
What do you decide to go after? What are your goals both on a short and long term?
Talking about short term one of the key things is customer service. We started with a pilot in the UK and it’s expanding now. And those are key pieces. You want to focus on use cases that you know that can bring real value. Especially here in the valley you can’t get around if there’s no money involved. We focused on return, payments etc. which are the very core drivers. And through chatbots we have seen people rate much better this experience rather than going through traditional channels. But there are always instances where the bot cannot completely take care of everything we did want or there are more complicated issues that the bot couldn’t solve. And those are areas where we need to interaction of a human assistant that can figure the issues out. Interestingly in the traditional channels we were not able to follow up previous discussions that we were able to manage with the chatbot by going back in the conversation and sort of creating a transition and give better answers through the human customer service.
In 140 characters what is your take on the future of chatbots?
- Jay: conversational UI is inevitable. So you better start now.
- Marissa: in the future we will have all our own chatbots that will interact with eBay chatbot and Jay chatbot to figure out what the best restaurant to choose.
- Gail: I’d recommend to start working on it now. It doesn’t have to be perfect, it just needs to be out there.
- Youri: we have seen a huge potential in the conversational interfaces which allow a better communication on the customer side. It’s just a work in progress, but not adding to much.
- HP bot: challenges with managing content. How do you manage content at scale and not just focusing on a small pilot?
- Did you realize that are some new use cases that you didn’t realized through the traditional channels?
- What is the story behind regarding the personality? How did you end up giving the name to your chatbot?
The clear conclusions drawn by panelists Jay Vasudevan, Lead Product Manager #eBay; Youri Bebic, Head of Product and Innovation #MoneyGramInternational; Gail Seanor, Sr Director, Marketing Technology #TGIFridays and Marissa Chacko, Director of Consumer Products #Foursquare are that chatbots are essential to creating optimal customer engagement and to cementing customer loyalty. The degree to which a company’s chatbot can respond and engage in human terms will be directly reflected in revenues. The role that AI and customer data will continue to play in creating ever-more-intelligent chatbots has only begun. The chatbot frontier is certain to expand and companies that fail to leverage AI-driven customer data for chatbot support will be left behind.
ALSO, LISTEN AT THE PODCAST:
Join the AI Expo World Series of events in 2018 for a series starting in London, moving onto Amsterdam and returning to Santa Clara.
AI Expo Global — 18–19th April 2018, Olympia, London
AI Expo Europe — 27–28th June 2018, RAI, Amsterdam
AI Expo North America — 28–29th November 2018, Santa Clara, Silicon Valley