001 – 10 Years Experience with Bots – with Botego Founder Ekim Kaya
Ekim Kaya is the founder of Botego, a client business in the bot spaces since 2007. He is sharing his experiences with us today. Listen to this interview to learn more about the industry.
Links mentioned in this podcast.
Feel free to comment or reach out directly to Ekim ekim at botego.com
You can find the full transcript below.
The Chatbot Podcast
Or subscribe with your favorite app by using the address below
Hey guys, it’s Bálint from https://chatbottutorial.com and today I have a guest with me. He’s Ekim and he’s the writer of the successful book on bots entitled ‘Bot Business 101’.
-Hi Ekim, could you please introduce yourself in a couple of sentences for our listeners who don’t know you yet?
-Sure. My name is Ekim and I’m the founder of a bot company. I founded Botego in 2007, I’m an engineer by education and an entrepreneur by heart and it’s been 10 years now, so we started before Apple acquired Siri. I was following the team of Siri back then. When I was studying engineering in Istanbul Technical University, I used to work as a call centre agent and that’s when it hit me that there’s something wrong with customer service, because people keep calling you and ask the same questions over and over. There’s huge room for improvement in terms of efficiency because it’s costly, time consuming, annoying, not for the customers, but also for the agents and the business, so it’s not a sustainable operation for any business. And that’s why, you know, all American enterprises switch to India or other Eastern countries to provide customer service but then their customers didn’t really like it, they had to switch back to America, all the costs associated with the operation is a problem again. The game has changed right after the Facebook invested around this time last year and I’m just now enjoying the fact that I’ve been waiting for this for 10 years now.
-Thank you. So, if I understand correctly, you were working during your university years as a call agent.
-Right. It was just for a short time, no one wants to have a career as a call centre agent, that’s why there was a huge turnover but that’s when I realised that we can fix this with software.
-Did it influence your direction like after the college or was it always on your mind becoming more efficient with technology?
-Yes, I’m always interested in fixing things in terms of making them more efficient, not necessarily physical, because I never worked as an engineer but when you see that you can build a machine, a software, a system that can work by itself, it’s really satisfactory, it’s really fulfilling for me. The fact that things work for me as I sleep: bots, the software, gives me satisfaction and I always try to find shortcuts and increase efficiency wherever I can.
-Bots have a big history but didn’t get traction before. Do you think that acquiring Siri was significant for the development of bots?
-Well, when you look back, you can now consider it as an important milestone but it’s been so long time and Siri itself didn’t really take off for a long time. So I’m not sure that it really changed things Facebook having a stronger connection with the brands and it had more influence over developments of the bot ecosystem. Because, if you remember, TechCrunch somehow found out that Facebook will be announcing a bot platform on their F8 Developer Conference last year and they somehow got this information and leaked it. Right before the conference I remember Microsoft and others, maybe Line as well, they announced their bot platforms so it was kind of a rushed launch. I remember, we checked up their APIs and they were really simple, because I think they just put together something in a couple of days, so this shows the power of Facebook over any milestone that they set out to release, and I think it’s the first time that all tech giants: Google, Amazon, Microsoft, Facebook, they are all betting on the same thing. You know, Microsoft CEO said ‘bots will live in your apps’ and Google CEO described, explained their vision as an AI in general and bots as a subset of it, you know, Google Home, Amazon Echo are now in our homes already. So, Siri is the first real step improvements, I mean creation, let me say, all this ecosystem, but Facebook changed things a lot more than Apple.
-I definitely agree with you and we probably wouldn’t be talking if Messenger bots didn’t exist.
-Right, you know, if you remember MSN, as a messaging system, Microsoft had MSN bot back in maybe 2009 or something, even earlier, no in 2007 we founded our company, so it should be 2006 the latest. So Encarta was their online encyclopedia and this MSN bot was offering content from Encarta. It was really simple, just based on keywords, it was bringing up the related article from the encyclopedia. This is the first bot that I ever used, mainstream on MSN. Years later it’s an industry now.
-So you founded your company in 2007. What was the first project you took on as a company?
-Well, first I went to my friends, who are founders of other start-ups and one of them is a bigger one in Turkey, Yemeksepeti. It was recently acquired by Delivery Hero for more than almost 700 million dollars, it’s one of the biggest acquisitions in the Turkish internet ecosystem and I told them ‘you have thousands of transactions every day and I can help you with your customer service interactions’. And they said ‘Ok, let’s try it but we don’t want to pay a huge amount for this, because we want to keep providing the human touch when it comes to offering help to our customers. So it could only have a limited use when people are asking about their passwords and those kind of simple questions, not about their orders’. So I said ‘ok, fair enough, let’s start’ and they’ve been our customers ever since. It’s been 10 years now. And then we went to an incumbent telecom operator and that was the first major client we landed and we had a lot of experience, we learned a lot from that project.
-What is the most difficult part on setting them the idea of chatbot or AI?
-That difficulty has transformed significantly over the course of 10 years, because in the beginning no one had an idea of bots, so we started our presentations by explaining what a bot is. Now most of the clients are coming to us, we receive a lot of involve, interest, so at this they have a basic understanding of what a bot has to offer. Back then it was nice to have, you could create a difference by offering your bot, because we are talking about before smartphones, it’s only web-based and websites, I mean, since there is no smartphones, the website usage is much higher compared to today, so it was really primitive compared to all these messaging platforms we have now. The reason why bots are really strong now is that these messaging apps are already in everyone’s mobile devices. Back then, the traffic on our bots was really, really low, so it was difficult to convince businesses to pay a decent amount of service fees. That’s why we struggled in the beginning to survive as a bootstrapping start-up, but now we just need to convince that you want to build a bot with an experienced bot developer, because it’s as easy as building an app now with all these platforms and all these tens of thousands of bot developers out there. Businesses don’t know who to start a project with; every time to convince them that it’s now a matter of experience when it comes to building a successful experience in time management, creating the right sequence, the decision tree. These are the things that they need to pay attention to. So, bot building is not rocket science especially after Microsoft’s bot platform and Facebook’s API, it’s a piece of cake. But you need to really offer a good experience, you know, not to disappoint the user.
-I see. So as I understand, back then it was more like talking about bots what are they what’s going on and now it’s more like ok, it’s probably obvious that someone will need a bot, but what is the best use case, what is the best company to develop it, right?
-Right, the use case is really important because they start with the wrong. They say ‘ok, how can we build a bot?’ This is the wrong question. I wrote a blog post on our official blog and I said ‘Don’t build a bot just for the sake of building a bot, that’s not the right position you want to be. You just have some pain points that bots can address and you can use our expertise to solve those problems for you using bots. They are just tools. When you’re basic in putting an IVR tool inside your customers’ mobile phones. So same principles apply when it comes to bot building but you have to have a problem to solve. I mean, what will your bot do? What’s your goal? What are your KPIs? You have to know them, otherwise you can’t measure your success rate, right? You’ll never know what you’re getting. And that’s the problem between mid-level executives of enterprises and top-level executives. Because usually top-level executives want to see results, that’s why they are up there, so you need to deliver. You need to convince the business units to do something meaningful and measure and keep improving all these.
-I see. I imagine bots were a bit different back then. What was your vision about bots back in 2007? What was your thinking in 10 years, so, for example this year, what would be a bot look like?
-Well, I don’t consider myself as a visionary in anyway, I just happened to start this business and I was lucky just to survive for 10 years. You know, just think that, as I mentioned it in the book, I was trying planning to shut down the company right before the announcement of Facebook bot platform, because I told it’s not going anywhere, it would never take off. I really didn’t have a vision about 10 years later, I just was trying to see the next step, and in terms of technology, we were really, I mean, when I look back now, it was a really bold move to try to build your own NLP engine because there are no platforms, there are no NLP engines available, no content databases. Now you can see that people started to adopt new job titles like bot designer, bot content developer, so now you have all these resources and it’s really easy. With trial and error we built our first tree versions of our natural language processing engine and as a local company we started in Turkish, which is a difficult language to deal with in terms of language structure, then we expended to other languages, starting with English. So I was again just trying to see the next step, I mean, I didn’t know if the company would survive for another six months because our costs were really minimal, that’s why we were able to survive, but we didn’t really have a real revenue for a long time.
-I see, so it was always about like getting a new client or a new project going on, probably.
-Yes, and it’s really so time consuming with the clients because enterprise clients it takes six months to two years to close a deal. It’s amazing how slow they can be. That’s the most difficult part of trying to survive as a bot start-up.
-Yes and I really enjoyed that you were talking about that in your new book as well, like giving practical tips about these clients and different types of companies and industries.
-That’s really important, because if you never sold anything to enterprises, you will always be optimistic until you face the reality. It’s really ugly. You should be really prepared to not get paid for a year, a minimum of 12 months. Beginning with the initial contact or your prospecting efforts all the way to signing the contract, invoicing, and getting paid, it takes really ages.
-I think it’s probably easier now with the Facebook announcement, I mean, back then there were probably some people who wanted bots to be more powerful, but they were, so to say, injured, like they were dissatisfied with the current state and now they have this story from back then, probably, but still the opening of Facebook might change their mind…
-Well, one of the most important chapters of my book is ‘Managing expectations’, because all these tech giants set the expectation bar too high with another line of products, AI products, and they have these amazing resources of information that they can process, I’m talking about about Siri, Google Home, Amazon Echo, they have all these data sets that they can use and millions of people using these products. So, what they are trying to issue with this technology is completely different from what we are offering as bot developers. I always try to get my messaging calls when it comes to explaining what bots are. We have a limited content domain that the bot is responsible for, which is your products and services. Don’t expect them to die into discussions of philosophy or even answer a question about what the weather is like today, because that’s not what bots are built for, if you’re talking about enterprise bots. People when they see a bot they feel like it’s a challenge to their intelligence and their first reaction usually is to test the limits of the bot and let’s see if I will be able to fail that; of course you will be able to fail that. Do you ever call a customer service number to try to fail a live agent? It doesn’t make sense, right? Even live people are expected to follow their written script, you can’t just call someone and ask something completely irrelevant. But again, because of these technologies, people are expecting a lot from bot, so our first purpose as bot developers should be to explain the difference between this strong AI and a practical application, which is bots that will answer customer questions. I always use the analogy of call centre agents. If you don’t expect something from the call centre agent, don’t expect it from the bot either. Another important thing is that there is always a time frame for improvement, for every bot project, because it’s always beginning from a scratch, you’re basic in feeling a knowledge base and it’s always the unique database, a set of questions, so it would never be the same compared to their first day and their second or third month. There’s always a time frame required to improve the bot, so they should be ready for this and they should invest the time and effort into bot development in this one to three months’ time frame.
-Yeah, it’s definitely always that you have this idea how it will work and then confronting with the real clients and then oh, I never thought to want that…So you have to update accordingly. like from the real feedback, not just like you and your friends.
-Yes, and we build model for the bot projects and we had 40 disappointments in the beginning, because when the customer realises that they need to keep entering new questions, the first day is almost always full of disappointment and you never get used to this. You don’t want them to test your product because their tests are so different from the real use case, the real customer never asks the questions that they use during their tests. This is what we’re trying to explain to them. They always come up with so formal and long written questions and we know for sure, that no one will ask that question, so we always tell them ‘it’s a numbers’ game, you need to cover 80% of the questions from day one, and you already received those questions through your e-mail, your support tickets or your call centre; so, let’s try to cover them first, then you will see that in time it will improve, of course you just need to keep entering answers. They are horrified when they see that day one is a complete disappointment.
-Would you say that because some of the limitations people would treat chatbots like a query, like a search in a way, because it is easier for the bot to understand the keywords?
-Well, there are various approaches when it comes to handling a question; some of them work like search, so the fact that there is no standardization in bot development is kind of a problem by itself, people don’t know what to expect from bots and they always have to try and see how that specific bot operates. That’s ok, that’s the natural evolution of the technology. Until there were two major operating systems in mobile that was the case as well, we had BlackBerry, we had all these other Linux based operating systems, and now there are two major ones and people know what to expect from Android and what to expect from iOS. I think we will see this as a similar process with bots, so the most important thing is to communicate up-front with the user what is bot capable of doing and what’s the content domain. As long as you say ‘I’m here to help you with this this and this’ and once the bot fails, you will again cover that, you will address that situation properly and then you will be fine. You don’t want to say ‘Hi, I’m a bot, how can I help you?’ People don’t know what to ask, they don’t know what they don’t know. So as a bot developer is your duty to handle the user experience from the greeting all the way to the end.
-Yeah, I agree definitely. There is a Facebook group where you are used to all the questions coming up again and again, so, the same way as a call agent, you need substructure, and that’s when you decided to write your book. Was it like that?
-Yes, exactly. I see how people are struggling finding customers or finding the right technology. Actually I didn’t cover the technology part, because there are millions of these sources who do that and I’m not a developer myself, although I’m an engineer, that’s why I think I would be able to provide more value by covering the business angle, because unless you have real hands-on experience with trying to sell bots, there’s no way you can learn this experience. So even though the things have changed over the course of 10 years, in terms of selling bots, I think I have a good understanding of what should be expected when you’re dealing with enterprise clients, so I wanted to share my experience and I thought it would be useful, I hope it is. I get good feedbacks, that’s why I’m happy that I did. One thing that surprises me is that people thought, I call them B2C bots, the individual bots is the way to go. I think that’s not the case, because people don’t really know what bots are yet, I mean, I’m talking about masses. If you stop anyone in the street, no matter what education they have, how old they are, they will know what a mobile app is. That’s not the case with bots yet. We still have time to get there and the enterprise bots, enterprise companies are the ones that will enable this, I mean, fundamental shift in perception of bots. That’s why now it’s time to invest in developing those bots and that’s where the money is. You see on this Facebook group, except for just a couple of personal bots, they are not really taking off. There are millions of personal bots and people just don’t know that they exist, right? But if you build a bot for Bank of America or General Motors, they have millions of customers and they will push it, they will aggressively market it just to get their bots to be used. So I think by the end of 2018 almost all enterprise client will have some kind of bots and after that it will be a good time to focus on B2C bots because people will know basically what bots are.
-What was the most exciting part about writing the book?
-I really tried to release the book as soon as possible because I saw the need to address most of those questions and last time I checked, it’s been awhile, I sold more than 2.000 copies so I hope these 2.000 plus bot developers are going through each chapter. I’m not sure if they are all targeting the enterprise space, but I just wanted to be the first person to answer these questions. There are a handful of other bot developers out there and they are positioned as some major companies like Next IT or Nuance, they required a Virtuoz, a French start-up, so I don’t think they are interested in providing guidelines to the bot ecosystem but I somehow felt responsible to do so. So just seeing all the positive feedback from the community really was the drive that I needed to write the book in such a short time. It took me less than two months to write the whole thing.
-Yeah, congratulations on the sells numbers…
-I think you had part of like protecting the whole bot industry in a way, so people won’t sell expectations and then deliver poor results. So in a way like protecting the whole system so that not so many companies would have bad experience from the start.
-Well, this is my message but of course I can’t just everything solemnly based on my own reach; people just need to try and see for themselves unless they read the book and decide to follow my suggestions. So some of that disappointment is inevitable, but you know this curve when it comes to adoption of any new technology; we have already passed the high point and now the expectations will be set, so we will be more comfortable in terms of what customers and users will expect from us and what we can deliver, of course. Facebook and Microsoft keep investing in the platform and that’s great, we will be able to offer better bot experiences. And bot development has already become a commodity, so we just need to create difference by offering a better experience now.
-Like how you addressed in your book as well and have many online points to have a way to update your book. So you addresses that the bot scene is very dynamic and changing all the time, and that expect now with all the big companies that it even accelerates in a way, right?
-Yes. Well, we receive at least five inbound requests for building a proof of concept for enterprise clients per week now and it’s a really strong indication of enterprise bots that they will become an industry standard real soon. And I think WhatsApp will be a game changer when they will offer an API that can be used to build a bot, they do offer now, but it’s not commercially available. So we are not suggesting to build bots on WhatsApp now, but once they do, I mean, my parents don’t use Facebook Messenger, but they do use WhatsApp, so I can tell that this is the way to increase our reach to a whole new level, then we will have hundreds of millions of new users, potential users, once that happens. I’m excited to see when that will happen.
-What do you think, talking about WhatsApp, what is the reason there is so much of silence from Facebook site? If I recall correctly that at the F8 conference this year last week there was nothing mentioned about WhatsApp really. Do they want to test things with Messenger first and then move to WhatsApp or what’s the logic behind it?
-Chris Messina was holding an online event and I asked this question to him and the answer I got was that what’s a bot is designed for is to offer point-to-point communication and you know now encryptions, communication security, is an important topic for them. So bots being in the middle of this communication requires some major changes in their structure. That’s why they were not able to adapt so quickly, but I think it’s a work in progress and they’re already testing with some selected businesses to offer customer service over WhatsApp. So, I think bots will be the next step for them and I think by 2017 we will for sure see WhatsApp being involved in the bot space.
-What do you think about God bots? Do you think they will be like this dominant factor and little bots will be connected to them or other the brands pushing the individual bots out.
-Well, anything can happen. It all depends on what the big players will decide to bet on. Currently it looks like it’s driven by individual brands and that’s not necessarily a bad thing. In the long run we will definitely see some platforms that will summon different resources of information. Again, 6 months is a really long time when it comes to changes, improvements in the bots space, I can’t even foresee 6 months later, so I’m not good in projections.
-One of the things I’m the most excited about is voice recognition and how it improved in the last couple of years and it allies well with the God bot idea for we as like we’re walking in the room, asking about the weather, texting a friend, asking about your marketing campaigns or results… So, do you think on the long run, voice will be bigger than texting or it will be more like balanced?
-I’m not sure about Voice. I was one of the early testers of Google Glass and the reason why Google Glass wasn’t a big thing like Google expected, was the fact that people don’t want to speak in the street with a device at this for now, unless it changes. It’s difficult to expect people to talk to themselves, at this point, to look like they are talking to themselves. Their exception is Google Home, Alexa, Echo, those devices are sitting in our homes or our offices, so it’s ok to use them. But it has some restrictions and important ones, so you shouldn’t have a huge accent for them to understand you, you shouldn’t be in a loud space, so of course technology is always improving, but I don’t think we will see them used as part of the bot technology in the time soon. So there are still things to be fixed in terms of speech to text, there are problems that need to be solved.
-I can definitely relate the accent thing, you should probably see some of the conversations I make with Google now.
-You know, Americans have a hard time trying to tell what they want to do on these automated IVR tools, they get annoyed, so yes, they should really be working well if that’s going to become a worldwide thing.
-Yes. There is an idea in your book I want to quote and it seems like it could be like the bigger next step after we have this voice recognition already solved. So you said: ‘One day you will have to explain to your child that you used to call customer service representatives manually on a smartphone and your child will look at you with that amazed expression and ask: so you didn’t have global net gene back then?’
-Yeah…I can tell because I have a 3-year old son and he is always watching this YouTube Kids app, the cartoons on my smartphone and when we go downstairs with the elevator, when the internet connection disconnects for just 30 seconds, I see how he’s outraged, you can’t expect him to not be online for even 30 seconds, because he doesn’t know what being offline is, he is just 3 years old. So, that’s why they are not patient, they don’t need to be, so I am 40 years old and even I don’t want to call a call centre. The idea of waiting in the line for 2-3 minutes is so annoying, so yes, it seems we already have these tools, the hardware, the software already available, it’s time to offer the service widely available to everyone. There’s a recent research about Millennials, they say that 68% of them prefer companies, brands that offer a self-service tool and they don’t want to call any real person if they have a tool to take care of their problems by themselves and they’re not brand loyal. You cannot expect them to be loyal to you anymore. They are experience loyal, that’s why you always have to improve what you have to offer and the best experience seems to be self-service now, because it doesn’t make sense to wait to speak to a real person when you can do it by yourself in 5 seconds.
-Yes, as a Millennial I can definitely relate to bot, so no brand loyalty and no waiting if I can solve it by myself.
-Same for me. I mean, it’s a shift, it’s changing, because the expectations are higher now and we have to deliver or we will be extinct, you know, what kind of troubles these big brands are getting into, just because they are late to catch the train just a couple of years.
-Yeah, exactly. So do you think that technology will be some partly like evolve with the humanity as well and then just become as a common and no entity at some point, or what’s this global engine thing?
-Well, yes. We will all have virtual assistance as part of our lives. Currently, there are different tools to handle different tasks but why not just one assistant to cover everything for you? I think we are not far away from that and it should already have been available. So, it’s like…now it doesn’t make sense to you to use ten different apps for different tasks. I think the Eastern countries already realise that; in Japan, in China there are these super apps that do everything in just one app. You can pay, you can chat, you can set meetings, anything you can imagine, you can do with just one app. The same applies for virtual assistance, you don’t need to have one to set your meetings and one to order food, another one to book your flights and hotel room. So someone will build a technology layer that will bing together all these basic data contents, so as a provider you will be able to be a part of this and you will be offering, you will be a part of an offering that has many aspects through just one smart assistant, rather than having 5-10 different ones.
-Ok, so it seems like I would have an assistant and the assistant would take care of all the communications for me, in a way.
-Yes, because you don’t need to know what works on the background, right? You just want to have one point of contact and the technology should take care of the rest of all these web services, APIs are already available and it should be presented from just one contact.
-Amazing, thank you. As a last question, I would like to ask you who you would nominate as our next guest to interview and why? We talk about chatbots AI and chat marketing.
-Ok. I would strongly suggest Chris Messina as a candidate for his really good understanding of the technology and he worked for Google, for Uber and he’s one of the opinion leaders now and I hope he accepts, because it’s really enjoyable to listen to him and he’s constantly learning, constantly sharing information, so I appreciate his efforts to improve the ecosystem.
-I will definitely invite him, I also listen to him as many occasions as I can or read his writings, they’re always of good quality.
-Tell him I said him hi 🙂
-Thank you very much Ekim. Where can people find you if they don’t have your book yet, or what about your company?
-Ok. So the company is Botego – a short for a bot with an ‘ego’ because it always improves and my email is email@example.com Ekim is an opposite of Mike if you have a trouble with spelling. Feel free to ask anything you want.
-Ok, thank you very much and if you haven’t checked it out, ‘Bot Business 101’ is the book and thank you very much.
-Thank you for having me. I have a shortcut for the Amazon URL, it’s http://bit.ly/botbusiness101, it will take you to the profile page of Amazon.
-Ok, some extra data for Ekim. Thank you. Bye-bye!
-Thank you, bye!