When starting to work on a chatbot, one of the first things we do is create a bot persona. After all, the bot persona is one of the most important attributes of a chatbot. Actually, without a clear, consistent and coherent personality, it would be harder for the users – and hopefully the customers – to trust the virtual conversational agent.
So, thinking about bot persona creators for Taking Turns, we needed to find a good one. How about a conversation designer that created the bot persona for one of the biggest fashion and shoes brands in the world, Adidas? Now it’s the time to meet Esha Metiary from Leiden, Netherlands, and from Crossphase – she’s the person behind Adibot.
In the 35th episode of our series, she’s talking about localization and multicultural aspects of bot creating. Furtherly, she even made history as the first guest whose text was had to be “beeped” (don’t blame her, it’s all because of a rude user). Let’s go!
How did you become a conversation designer?
First of all, I have a short background in customer support. After working for an online auction platform, I wanted to go away from that. So I interviewed for a couple of positions that had to do with copywriting or content management.
Actually, at the current company I’m at, CrossPhase, I applied for a position as a content manager. However, during the interview, it kind of became clear that my qualities were more suited for a conversation design position. So one of the recruiters said during the interview – ‘well, I think it’s time we go over to Plan B’. And if there’s something you don’t want to hear during the interview, it’s ‘Plan B’. However, they introduced me to the position of conversation designer, and I had no idea what it actually was. So I had to read some things. Finally I thought, well, this is pretty interesting, and I’ve never done this before, so, yeah, let’s do it.
What skills did you have that made it so appealing to you?
So the basics for me were the background in customer support and the basic interest in technology, of course. When combining those two, and adding little psychology, you have a perfect recipe for conversation designers.
What is the bot or project that you’re most proud of?
Actually, I’ve worked on a couple of projects in the 1.5-2 years, and each project teaches you something new. So I’m quite proud of all of them. However, I think that the one that I’m most proud of is the Adidas one, which I’m currently at. Basically, we started to build a bot from the ground up. Although there was some groundwork done already, we had to start anew. New platform, new intents, new dialogues. So we started with building the bot persona, which was, I love building personas.
After that, we had to connect with so many different departments – legal, brand communications, customer support. So many people were actually on board in this project or are actually on board in this project. That makes it the project I’m most proud of.
Since you’ve mentioned it, what’s the secret to building a good bot persona?
Well, I know that the opinions differ there because you have to balance the company needs versus the user needs. Yet, I think that a chatbot is really an ambassador of your brand. So stick to your brand values and make them also clear to the customer or the user. Thus, if you have a clear defined chatbot with core characteristics and a very defined way of speaking, I think that is the most important part of creating a good bot persona.
Previously on Taking Turns | Watch the whole playlist
Were you ever asked to change course or fix things in the bot persona, during or after launch?
Yeah, this immediately makes me think of a principle in copywriting, which is ‘killing your darlings’. So the bot persona for Adidas, which is Adibot. Initially, we made her a model, a bit of Beyonce. Because Adidas has the Ivy park line, and she’s just such a role model for everyone, all women, of course. Then we found out that Ada, as she was called back then, is actually a bot service online already. So that was one of the first things we had to change.
Also, there was another major change we’ve made. A principle in bot personas is that they don’t say sorry too much. And I designed this with an English screenwriter and myself, I’m Dutch. Since I’m not really sure if you’re familiar with Dutch people, they can be blunt, short and unapologetic. Surely it doesn’t mean that your bot has to be this way. Yet, there is quite a cultural difference if you apply those principles from a Dutch copywriter or Dutch conversation designer into an American bot. Because that is what our core business for Adibot is. The American market desires much more apologies – saying ‘sorry for the inconvenience’, sorry for this, sorry for that. Although that was not really a shock for me, it was something that I had to… get used to.
So while we initially designed the bot to be not so very apologetic, we had to come back on that decision. Then, we added a little bit more apologies and made it a little softer for the public.
What’s the most important thing in a chatbot, In your opinion? It can be in the bot persona as well.
In my opinion, as I said – chatbots are brand ambassadors, which means it has their values, their core characteristics. Also, we should help our customers as clear as concisely and direct as possible, because customers are not waiting for long answers and beating around the bush. So it’s very important to balance between the core characteristics of your chatbot and thus your brand, and not bend to the will of the customer too much. In this sense, it’s very important to manage expectations. Not only tell the customer what your bot cannot and can do, but also what it will and will not do.
One of your core principles in your work is inclusivity. So how do you make a chatbot inclusive?
So textual, it’s quite easy to make a bot inclusive. The basics are – prevent pronouns but use plural pronouns. Also, realize that as a conversation designer/AI trainer, you don’t have all the knowledge that is needed to make a bot inclusive. So make sure that you connect with all kinds of people who know things about the customers who come from different demographic backgrounds. Then, take the knowledge that the people besides you have and apply it to your own work.
Can you recall funny, amazing or awkward stuff that happened to you designing the chatbot?
Oh, yeah, I do. Pardon my French but small disclaimer – one of my favorite things to do is actually go through all the things that customers say to us. Also, as you may know, one of the biggest products that Adidas sells is Yeezy shoes. Then, these products are dropped through a special process. Hence, it’s always very busy around that time. People don’t always get what they want, because there’s limited stock. It’s exclusive access, etc etc.
So to cater to the needs of those customers, we go through the sentences of the things they say and filter out training phrases. This way, we can feed that back to the bot. And here it comes: one of the sentences that an actual customer said was, ‘Whose D*** do I have to suck to get a pair of Yeezies around here’?
Of course, the customer is very frustrated and we get that. And it’s really sad to see them not getting the pair of Yeezies they want. However, seeing such a sentence in the training phrases is just brilliant. Naturally, the bot didn’t understand the sentence that the customer said. So he said to the customer, ׳Sorry, I don’t understand׳. Then, the reply the customer gave after that was, ‘I didn’t stutter’. So, yeah, it was quite a funny scenario.
Can you tell me more about the company you’re working for, Crossphase? About your daily routine, your work there?
Yeah, yeah! CrossPhase is a company that specializes in online communications services. So my colleagues are content managers, copywriters, and of course, conversation designers. Everything that a website could need to improve its services, textual or website voice – we can offer that. Here, I’m one of the conversation designers. Internally I usually keep busy, of course, with my Adidas project. Then, we also have sessions to improve our service or brainstorm about where conversation design is going and what Crossphase should do with it.
If somebody wants to become a conversation designer – you had it like Plan B, If somebody wants to do it like Plan A, what would you recommend?
Yeah, there are a couple of things you should do. Of course, there are always online courses. Also, you can train in UX or customer support or do free trials of bot platforms. Anyway, the thing is – just get started. Because I’ve noticed that the community around conversation designers is very happy, little and welcoming. So just send out invites to people industry already, connect with them and ask them, hey, do you have any tips? Be ballsy. Don’t be afraid. Just do it.
Can you give us one forecast for the future of Conversational AI?
Just one? First, with the bots becoming more omnichannel, in the future, I will also see bots using biometric identification. So in most service bots or retail bots, customers are identified now by their customer number, order number, or address. However, in the future I see bots that will identify users with their fingerprint or with a facial scan, to bypass any other details that the customer should possibly give, which can frustrate the process.
Another forecast is that chatbots will become much, much, MUCH more Proactive. Nowadays they are reactive – you ask the question, we give the answer. In the future I think that chatbots will actively engage with the customer. For example, in Adidas, I will see it happening that we see that a customer is very interested in a certain line of products. Then, when a new product is bound to come out, the chatbot will approach the customer. It will say ‘hey, I see that you’re purchasing a lot of these products. We’re dropping the new line in a few seconds or in a few months. Do you want to be updated?’
Also, it can happen with telecom companies. The chatbot will say, ‘hey, I see that your subscription is almost up. Do you want to see what kind of offer we have for you to renew that subscription?’. I really see that happening.