Transcript of the podcast "Let's talk about IT" # 035 from May 15, 2019. Subject: Chatbots.
Krzysztof Kempiński, founder of the "Let's talk about IT" podcast: Today's guest is the founder of the Poznań Software House High Solutions. Programmer with 11 years of experience. CTO for many years. Now the CEO. Passionate about the Laravel framework and creating Chatbots. He runs the blog Turkusowy prezes. Speaker at conferences related to programming and turquoise management. In his free time he runs and trains crossfit. Adam Matysiak is mine and your guest.
Krzysztof: Hi Adam! I am very pleased to welcome you. I am glad that we will talk about a topic that lies within your interests - chatbots.
Adam: Hi Krzysiek. Thank you very much for the invitation. I hope that we will talk about some very interesting things about chatbots.
Recently, no episode of my podcast can go without a standard question. Do you listen to podcasts and if so, what are the most common?
I prepared myself for this question. First, let me say that I'm more a book than a podcast person. I also listen to podcasts. I have been listening to your podcast for a long time. And of course, Business in IT is two of the most interesting podcasts for me. Ones related to management are Mała Wielka Firma or Manager Plus. Plus, of course, some technical, foreign ones, but from the Polish backyard are the ones I mentioned.
Thanks for listening. The other podcasts are also very good choices. Perhaps we will start by saying what chatbots there really are and what types can we mention?
Wikipedia says that chatbot is a computer program with a conversational interface. So we have something that communicates with us through conversation. This is a way to automate repetitive activities. Most often we use it to serve the customer who looks for information. This basically can be used for everything. The idea is to replace the man on the other side of the chat. As for the types, they can be divided into several ones.
First of all, the first division were the script chatbots - that is, three options to choose from and the user chooses. We communicate via chat, but it is fully automated. The NLP chatbots - Natural Language Processing, means those that understand what we write, are becoming very popular. They guess the intention and react accordingly. The second big division are dedicated chatbots, in which we specialize. We create a solution from scratch. There are also configurable chatbots, that we can take care of. This is also a quite popular solution. Another type of division is chatbots from customer service or from increasing sales. The main division is scenario, NLP, dedicated, configurable.
Recently, there is a buzz around Chatbots. It's both technical and strictly business. As HighSolutions, you provide chatbot creation services. I wanted to ask you what types of companies are your clients? What industries do your clients come from?
When it comes to industries banking is the most active. Also e-commerce. I think that also production companies. These are the industries on the top that are investing in chatbots and I think there will be more. Big famous banks, chain stores and production companies want to automate the most. We see more and more on the market that smaller companies are also interested in chatbots. They are looking for ways to automate and improve the quality of customer service.
Do you think that the advantage of large companies results from the lack of awareness of smaller players that they can gain something from it?
I do not know. I am not sure if this is a matter of consciousness. I think that large corporations also do not have full awareness. We've been working in chatbot field for two years and only now can we see that the market has started to get interested in it. I think that is also a question of funds. Large corporations can afford innovations and tests, and small ones have to focus more on everyday life. This is not an advantage. Maybe more awareness of the marketing department.
We have an implementation for the Aquapark in Wrocław. This is a multi-million dollar business. A relatively small company compared to corporations, but they sensed that they need support for the customer service department, especially during holidays, when a lot is happening. Here, it was definitely worth implementing. They are very happy, as far as I know. This is an example of a relatively small company that has entered this trend very strongly. I think that chatbots are for every industry where we have customer service and it does not have to be a big implementation. For 10,000$ you can put a chatbot, which will automate and improve.
It's good that you see the trend of chatbots spreading among smaller companies. This means that this topic is becoming more popular. What value does the chatbot deliver to the company that orders such chatbot and for the customers of this company?
It depends. It depends on the needs of the industry or the company. First of all, this is where we now use chatbots, it is a relief for customer service. We have a lot of cases where the same thing always happens: the same repetitive questions that we waste time on. It is: checking the status of the application, the status of the package, and this is exactly what you can automate. That we, as consultants, could deal with something more demanding, that is a customer who has a real problem, not two simple questions. This is the automation of customer service and this is the greatest value.
For large companies it is a good PR operation, it shows that they are modern, they make innovations, they work hard to provide exquisite customer service. What value? Chatbot operates 24 hours 7 days a week. He does not get tired. It can handle many people at once. As we add machine learning to it, it can be better and better. These are the most tangible values that the company's chat provides.
Do you want to talk about chatbot for your company? Click here and write to us!
Let's go to the “meat”, because this podcast is mainly listened to by people who deal with technical matters. I wanted to ask you what the process of creating a chatbot looks like? I mean meeting with the client, analysis of his needs, internal design, chatbot software and implementation. How does this flow look like?
From our perspective chatbot is a normal IT project, only with a different interface. For us, it is an IT project that has several phases as usual. The first phase is the analysis, i.e. we wonder why this chatbot is being implemented. For whom? Who will be a user? What are the business needs? What are the current inquiries? What intentions are behind their questions? How to communicate - formally or informally? We create a chatbot person. This analysis must be in-depth. Sometimes we tell the client that he does not need a chatbot but some other IT. Sometimes we advise against this solution when we see that it does not make sense. This phase is always needed in any IT project.
Then there is the design phase, that is, the creation of communication flow. How do we talk? How is this conversation going to take place? How do we respond? Which language? How do we handle the exceptions? What happens when an error occurs in the style of misunderstanding the answer? Then copywriting is also important to make this communication natural.
Then there is the programming phase. We program chatbot and here we have machine learning, or labeling to determine what intention was behind the question?
Then it's implementation phase. Most often this is on Messenger, so we report to Facebook. It takes a while. It's not that we report and we leave. This is a bit of a hurdle. Facebook checks if this chatbot is working correctly. Only then It can then be put on the Facebook page.
The last phase is maintenance. We analyze what users enter. We monitor. We train when someone asks a different question, and sometimes you have to add a new intention.
These are the basic phases: there are 5 of them, As in any IT project from our point of view.
Let's go a step further. Are certain programming languages, libraries and external services used to create chatbots? What do you use every day when it comes to tools?
When it comes to programming languages, I think that you can use any language. There are many solutions, be it in Python, Node.js or PHP. This is server communication, so it is not a great philosophy. More interesting things happen when we want to add machine learning. External tools are the most commonly used here. For example, Dialogflow, which belongs to Google, or Wit.AI. Both of these solutions help us implement machine learning to recognize the intentions of users who write to us. I think that these are the two most well-known external services that you use. I often have libraries in Python and Ruby that allow us to create chatbots. I think that there are solutions in every language. We can use what we know and add a library.
Chatbot is a normal IT project. It differs only in the interface.
Any services that come to your head that you use? You mentioned Dialogflow. This is probably the main player. Are there any competitive solutions that can be used?
We also use Wit.AI a lot. This was the first solution that served Polish language. There is also the NLU Race. I think there are several of these solutions. There are also other solutions that improve things, but I think that to start with, it's worth taking an interest in the topic of NLP to get into the topic of machine learning. Train bots, not just program them. From such very well-known solutions to create chatbots, and not to program them from scratch, is Chatfuel. It is the most popular. We can accurately determine the conversation exactly and immediately implement it on Messenger. There is this problem - it is a little bit harder to do integration with other websites and communicate with the API. We also use Botsociety, which allows you to block conversations. This is a cool interactive demo that we show to the customer. You can immediately create a bot and connect it to Messenger. Our projects, however, always have integration with external systems, so we build it ourselves. We use these solutions to show the customer how this communication might look
You said that in Chatfuel you can think of such a chatbot. I wanted to pull this thread, because we live in a time where there are a lot of SaSS services in the cloud, where it is possible to set up and sort out many things. Do you have to be a programmer to make a chatbot?
I think not. First of all, we have Chatfuel, which allows us to create a bot. It's like WordPress. Then there will be simple solutions that will not allow us to strongly integrate with external systems. I think that here is a great value in automation. If we wanted to put up a simple chatbot, yes. You do not need to be a programmer, but it will not be as advanced a project as if we were designing it from scratch. On the other hand, it can be done in one day. The biggest fan could create his external project from scratch.
Let's say that we are behind the design phase, behind the chatbot programming phase. I wanted to ask you - how is this solution tested? This input, which is what the user can enter into the solution, is theoretically diverse. It's hard to predict what users can enter. The human imagination knows no limits. Does this affect how the chatbots are tested?
I can agree with you that the human imagination knows no limits. The most common question is: "What is the weather like?". Often, we immediately implement this question to show that our chatbot is not stupid. This is a bit of a problem. You can test in different ways. We do it in test unit. That's why we have all NLP to train him. Thanks to machine learning, we can guess what the intention is and we are able to handle something that previously chatbot could not. If we used Deep Learning, we can do it much better. You can not predict everything. Users will always ask other questions or make so many mistakes that our NLP model will not guess. We need to be able to answer that we are not able to handle it and in such a way that the client would say that our chatbot is not flawed, but we asked a bad question that nobody had asked before. This is a difficult issue.
I think that it will change a bit in the long run. It is also the consciousness of people, where everyone thinks that chatbot is artificial intelligence, so we can talk to him as with a human being. There are, of course, chatbots that are able to pass the Turing test, but these are not ordinary chatbots that are used in business. The normal bot will not be able to answer a lot of things, or it should not. What is the answer to the question of who will win the presidential election? This is not a question. It is also a matter of market awareness.
There is often such a phenomenon where we recognize that we are talking with a machine, not a human, then we start writing a simple language, as for people from abroad. A very simple English language to understand. Sometimes they write commands and this also changes the way people talk. I think that ultimately we will not be able to recognize this difference - whether we talk with the machine or with a man. This is a matter of technological work, the language we will use, the universality of solutions. This is a difficult plot, so it takes some time to monitor conversions for a long time after implementation. Human imagination, however, can find a limit and we will be able to answer well at 98% :)
Should we use external platforms? Creating a chatbot for Messenger, we move in a space that does not depend on us. We have no influence on it. Is it, therefore, possible to run a chatbot on your own server or on your website?
There are issues related to GDPR. As we create a dedicated chatbot, we put it on our server. Everything that happens on the server side is ours. If we have it on Messenger, all communication goes through Facebook. We had an inquiry from one of the banks, how to make a chatbot to reset the password, we said that on Messenger we can not do it, because it will violate all the issues of the GDPR. We suggested creating a chatbot on the site. You can make a chatbot on the website. In principle, it is a livechat that is automated and works the same. We have much more control.
We also have implementations and examples, when the client needs to have a chatbot on the store's website, or a chatbot that shows around the store, and not only answers questions. Here you can also use it. We also have GDPR. Facebook has personal details, but if we were sending more, we already have a problem. You have to be careful. But you can put chatbots on a website or a mobile application. This is just a matter of interface. This is not a big problem. This is an interesting option, and not everyone is aware of it.
You said that e-commerce is one of the most common cases of chatbots. Let us assume that we have an online store with jewelry for women. What could I use in such a store? What benefits are here for me as an owner of such store?
For the online store with jewelry for women, we have made a test for the Polish brand. We wanted to show what their customer service looks like and how much they are losing. We've done the test. We wrote on their fanpage a message. We wrote to another foreign store that had a built-in chatbot. Then, we wrote that we are looking for a necklace for the fiancee. We got the answer after three hours, that you should go to the store and choose yourself there. A bit polite, but the intention is the same. In the foreign store, on the other hand, after 10 minutes or 6 minutes we had a recommended necklace with the "buy" option. We showed them that in 6 minutes they could do sales, and in Polish shop, they made no sales. I do not think they were bothered by it because they did not want to talk to us anymore :)
But this case strongly shows that you can automate many things, you can have cross-selling or up-selling solutions. Increase sales strongly, increase conversion significantly. It would definitely work for the online store - "what is the status of the shipment", "where to file a complaint". Different things that we need to handle, and all this is done by chatbot. We are more efficient or we have better customer service. This is a big plus. Large brands, because it is one of the largest brands in Poland, such top 3, are not aware of this. I think that it is still a lack of awareness, but I think that they will decide to do it one day. We had a perfect example of how chatbot can make us sell.
E-commerce is one of the industries that most often uses chatbots eg for up-selling and cross-selling.
You spoke about consciousness. I wanted to ask you how the chatbots market in Poland looks like compared to abroad? Awareness of companies, marketers, users who talk with chatbot. You entered a niche of chatbots. Looking back in time as a person who manages a company like yours, was it a good decision? Is this a perspective market?
I think that the foreign market is much ahead of us. The first time we met with chatbots in 2017. I just went to Brussels for two months. There were meetups about chatbots practically every month. In Poland, there were practically no companies that offered it. There were almost no implementations. We have found that you have to enter it. The Polish market is at an early stage mostly when it comes to the theory of the spread of innovation.
I think that now is the moment to enter it when we are still innovators. It may change soon. As the example of a jewelry company shows, market is not fully aware that something can be done. When it comes to business decisions, it was a good decision. We specialize in tailor-made solutions for business. This is one of our next services or examples, what we can do for business if we need to improve customer service.
I think that in a few years chatbots will be popular and universal solution. From our perspective, we must be happy that we do it, because we can talk with the largest companies in Poland. Whether the banking, FMCG or production industries. We like to cooperate with large companies. With small also, because these are different solutions and different satisfaction. When companies we know reach us, it is cool when we can talk to them and try to help..
Do you want to know more about what chatbot can do for your company? Click here and let’s have a chat!
There was a lot of talk about the GDPR. You have also touched this topic. Is placing a chatbot on the site subject to regulations related to the processing of personal data? And what it is like if you want to use an external platform like Facebook, Messenger?
The topic is quite complicated. I always recommend consulting a lawyer. There are no jokes here. If we would like to obtain very sensitive personal data, I think it is better to do it on your terms, i.e. on your website or in a mobile application. Not all of things should be sent by Messenger. You have to be careful and always have to consult with a lawyer or an IT company that has an experienced lawyer. This is also a solution. We may sometimes not have personal information in a chatbot. Then there is no risk. Then we process personal data and you have to be aware of it and take care of it accordingly.
At the beginning, you mentioned in conversation that you treat a chatbot design as a standard IT project, with roles that you can find as standard in such a project. On the other hand, we know and also confirmed that this industry is developing very fast. I would like to ask you if there are any professions or specialties that are specific and created only for the production of chatbots. Personally, I've heard about such roles as chatbot script designers. How does it look from your perspective?
I've heard about the bot's personality architect. I think that various interesting roles will now be created to show that we are modern or that our work is not standard. I think that it is still the responsibility of the marketing, digital or communication department. As part of this, they can specialize if someone has experience writing chatbot scripts or creating this persona, or bot. I do not think there will be such a demand. It may be more as a role or a task. I do not expect such names to come.
Data Scientist - this is a position that will be consolidated and will be needed further. In the case of bot, I think it is too small. It is more a communication specialist.
Let's look a bit into the future. What kind of development trend do you see when it comes to chatbots? On both from the technological and usability side. In which direction this industry is changing and what we can expect in the near future.
I think that a lot of things will happen. Certainly this understanding of the language must be at a higher level. We install the library and this understanding is at a very high level. For sure, it will go higher to make it easier to work and communicate.
Looking into the future, I believe that people should not be created for very repetitive work. I think we can do better. We will analyze everything that is repeatable so that we can release our creativity that we have. You will want to deal with interesting and new things, not repetitive ones.
Chatbots are one of the ways to automate this work. It is not necessarily something bad. We will have to deal with other things. Voicebots will probably come from such solutions. They will be better and we will be able to talk to Google Assistant without any problem.I think that technologically it will be possible to create it even simpler, to assemble it from blocks. We're going to make something out, we'll program something. It will be easier and artificial intelligence will be more important. For sure we will try to automate everything we can. This basic customer service is the first to do this. It will also be simpler for us. You will be able to deal with everything faster. But you must be aware that these chatbots will be everywhere among us, it will be natural in a few years for us.
Chatbots in the future will be a way to automate repetitive tasks.
Exactly. This is another block in this technological revolution that we observe and which we are a part of. I believe that during this conversation we have inspired at least a few listeners to deal with chatbots in the perspective of their professional career. Therefore, where to get knowledge from? How to learn? Where to gain experience when it comes to chatbots? What would you advise such people?
As a self-taught programmer, I will recommend doing research. What technologies are available? How can we create it? What are the differences? What are the differences between platforms on which chatbots can be made? We can choose something that interests us and what we know. Get there. In trying solutions, introducing novelties.
In Poland, Messenger has no control related to payments, but they are already in the US. If we test our novelties in the US, how they will come to Poland, we can be the first to do it. It's worth following what's going on abroad. Blogs, technology magazines, conferences. In Berlin there is a global Summit with chatbots every year. It is worth going to events, or come to a company that makes chatbots. This is the most practical experience. One can easily make a chatbot in any language, so you can have fun at any time.
Cool. Many thanks Adam. It was a really interesting, inspiring conversation. I learned a lot myself because with chatbots I did not have much to deal with. It's nice that we came up with this overview, from mundane issues related to programming language and technologies around chatbots, to futurology. It's good. Thanks a lot. I keep my fingers crossed for your speeches and various conferences in which you participate as a speaker. I also invite listeners to participate in these conferences and meetings. Thank you very much.
Also thank you. It was very nice to talk to you and share your knowledge about chatbots. I think people will be a little more aware of how everything works, why and what awaits us in the near future.
Krzysztof: Thank you so much. Bye!