What is the story behind CoSupport AI? How can AI improve customer experience? How did CoSupport's AI engineers obtain the patent for the technology? Read our new article and find the answers to these and other questions. We present the transcript of the interview between Cristian Dina, Alex Khoroshchak and Roman Lutsushyn, the CEO and CTO of CoSupport AI. You may find the podcast right here.
Tekpon Podcast With CEO - How CoSupport AI Transforms Customer Service
Cristian: First of all, I’ll let you present yourself and tell us more about the product.
Alex: Sure, I can start. My name is Alex; I am the CEO of co-support AI. Co-Support AI is basically a startup that develops and implements AI solutions in customer support.
Christian: What would you say is the biggest problem that you solve?
Alex: The biggest problem that we solve is definitely the continuously growing customer expectations. If we take sales for study, for example, approximately 80% of customers demand immediate responses, while 60% are not happy with the resolution time and time is crucial. In that regard, traditional support approaches are dying and that’s where CoSupport AI steps in. So, we offer AI solutions that, first of all, automate a portion of your support request, and then boost your support team’s productivity by providing them with ready-to-reply suggestions. Also, we have the BI product, business intelligence that can really convert your customer correspondence into revenue.
Cristian: Who would you say is the product for? What type of companies are you targeting? Who is the audience?
Alex: The audience is definitely companies that have at least 5,000 support requests monthly so this is from mid to large-size companies.
Cristian: Okay, in terms of features, what would you say is the most loved feature that you see in the product?
Alex: Very good question. So our pricing model is built in a way that all our subscription plans provide you with access to all the features right away. But I guess we’ll talk about pricing a bit later so, regarding the most popular feature, I think that that’s definitely the option to customize the AI models that we build for our clients. Because we build custom models and we can even influence the AI model behavior. Say you want your AI on the website or app to be more sales-oriented, to entice people to share their contact details or book meetings with your sales team. So, we can set up the model the way that it will do that. Tone of voice, also we can adapt your communication style so that’s something that our clients really like and appreciate.
Cristian: What are the use cases of the product?
Alex: Okay, there are actually a lot of different use cases. First of all, that’s the standard support, you have an app or you have a website and you want to have a widget on the app or website and the AI will resolve all the incoming support requests independently. So, that’s the first use case. Then the second one, which I mentioned a couple of minutes ago, the sales one. Also, use cases differ depending on the AI solutions that you would like to leverage because we offer different AI products. The first example is if you want to automate: that’s the AI agent that resolves support requests independently. So we can leverage your documentation, your public help center, and past correspondence and, with the help of the data sources, automate a portion of your support workflow. The most important part here is that our model does not hallucinate. This is a common issue with, for example, ChatGPT, and all the other solutions that are leveraging open AI technology. That's because we have our own unique AI architecture that has been recently patented in the US. That’s one use case. Then you may not trust AI yet. There are many companies that just don’t want to allow AI to handle support requests independently. In that case, we have a separate product that, instead of responding to customers independently, provides ready-to-go reply suggestions for your support team.
Cristian: Yes
Alex: They don’t need to spend time crafting responses from scratch. And on top of that, the AI business intelligence so we integrate with your communication channels CRM assistance, and then you can leverage all of the data to handle tasks like, okay, “What are the top five issues reported by our customers this week?” or “What are the top three feature requests of this year?” Usually, that’s a really time-consuming task because support agents need to manually track the data, then someone needs to analyze this data and create reports. But our AI handles such tasks in seconds.
Cristian: That’s amazing and thank you for sharing the use cases. Now I think it makes sense for everyone and in terms of the integrations which one is the most common to see being used by your customers?
Alex: Yeah that’s definitely the integrations with their CRM systems. so we have integrations with the most popular help desks like Zendesk, Freshdesk, Intercom, Zand oho Desk on top of that we also integrate with business messengers like Slack and Microsoft Teams and lastly we offer custom integrations so integrations with clients' in-house system may have some order management systems they built in-house and our AI can be integrated with that order management systems so that when people reach out with questions like, “Hey, what’s my order status?” instead of generating some default generic answer like, “Here is the link to the tracking page, enter your order ID and you’ll find the status of that” it’ll go to your backend, locate the order and come back with the actual order status. “The order is on the way to you. Expected delivery time is Monday 3 PM.”
Cost-Effectiveness as One of the Main CoSupport's AI Benefits
Cristian: Okay, and in terms of pricing, how does it work? Do you charge per response, per month? How does it work?
Alex: Yeah, our pricing consists of two components, the one-time setup fee and the monthly subscription. So, the one-time setup fee varies depending on the systems we’ll need to integrate with. But usually, it’s within $500-$5,000 and then the monthly subscription. We offer different subscription plans. The essential plan, for example, costs $2,000 and it includes 10,000 AI responses. By AI responses I mean the automatic responses in more than three languages, 24/7. Then the response suggestions are generated for the support agents and interactions with the business intelligence.
CoSupport’s Story: Becoming All-in-One AI Solution for Customer Support
Roman: Hi, yes, that’s time. Let me quickly introduce myself then I will continue with the company story. My name is Roman and I am the chief technology officer at CoSupport AI, so basically, in simple words, I am in charge of all technical stuff in the company. Also, I am the inventor of our core product CoSupport Agent so now let me quickly tell an interesting story about how our company was created. So, basically that was I guess October of 2020 and for that time I was hired as a regular machine learning engineer for a support outsourcing company, and I’d been working for a few months in that company. I created a new solution for how to automate ticket processing—time, how we can eventually optimize it and reduce delays before the response. So, after that, the CEO and investor from that company just had a conversation with me and said, “Hey Roman, it seems like that’s a pretty solid and cool solution so would you mind if we run a new company and you will be in charge of all technical stuff? You’ll be in charge of all improvements for that.” from that time we started our company. But I guess after two years, or maybe two and a half we started with the name of NLP Lab, which stands for Natural Language Processing Laboratory. So, we are the guys who know what to do with that stuff but with NLP. So, after that, we discussed and decided basically that we needed to move forward with an end solution not just the laboratory. I guess one year before, in the summer, we established CoSupport AI. So CoSupport is like your second co-pilot that will leverage your support like your customer experience in real so that’s basically it.
Cristian: So, it’s been four or five years now?
Roman: Yeah, so, that’s 10th October it’s gonna be four years.
Cristian: And you started with three co-founders, right?
Roman: Yes.
Cristian: And today, how big is it?
Roman: We have the same number of three cofounders, but our team is about 10 people 10 super, highly talented people super experienced people. This is led by Alex, our CEO. By the way, Alex has just unbelievable experience in customer support so that’s why he knows what he’s talking about. Also, one of our investors is Daria Leshchenko, CEO and managing partner in SupportYourApp, a huge company.
Cristian: I know. Daria has been on our podcast before.
Roman: Okay, cool.
What makes CoSupport Any Different From the Rest of the Companies?
Cristian: In terms of the market, I know there are many solutions starting to pop out right now in customer support in AI. What makes CoSupport any different from the rest and how do you see the market in terms of the competitors?
Alex: Good question. Well, we think we developed the best technology on the market to automate and streamline your customer support processes. Why? Two patents prove that we are unique. Uniqueness in the field of AI is definitely something that makes you stand out from the crowd because, as you mentioned, many new AI companies are founded. And you really need to have something special, something unique that makes your solution better than the others. In our case, that’s our patents. We have a patent in unique AI architecture already and an additional one is now pending. It’s about the data processing. So by having these two patents, we are sure that our solutions work faster than our competitors' solutions and that they deliver more accurate responses. For example, let’s say hallucinations. Hallucinations are a common issue for GPT and GPT-based models why because Chat GPT has access to almost the entire internet and since the information from different data sources is often contradictory, you never know which data source ChatGPT uses when it generates answers. That’s not the case with our solution, because we have our own unique architecture and we control what AI does and what knowledge base AI uses when it generates answers. So, when it cannot find the answer instead of making it up, it will just escalate it to your support team.
Patented AI Technology
Cristian: How do you get the US patent? I think it’s really interesting. How long does it take? How do you apply?
Roman: I can elaborate on that. So basically as Alex mentioned we have two patents. The first one is already covered for us, confirmed, and approved. The second one is pending but it’s a paid application. For the first one that’s a kind of system and systems and methods for generative language models architecture so basically that’s the one for a unique model architecture approach architecture to generate text generate responses. For getting that patent, I guess, we spent a few months of my research, a few months of my investigations, and the first implementations from my end. After that, we started our, you know, legal stuff, and that, I guess, took up to almost a year. The entire procedure took us up to one year. For the second patent basically I guess just within two or three months we gonna have the second one because we submitted it. We just wanted to move forward only with the one and now we just want to finish with the second one, so I guess in the nearest month we got to have the second one. The second one is also important because the one thing where we are different and better than our competitors is data privacy. So, the second patent ensures that we know and have our own approach to how to process the data and how to anonymize the data. We have our own models, pre-trained, and even model ensembles to pre-process, clean the data, and fully anonymize it. We can replace up to 16 entities like first name, last name, mobile phone email and so on and so forth. So, basically that’s really really important stuff because right now in our century data is one of the most important resources so we need to be aware of any kind of data leakage, personal data leakage, and so on.
Cristian: And would you say is this a big selling point to new customers today? Do they put a big price on this that you have this patent?
Alex and Roman: Yes.
Cristian: Definitely, yeah. Basically this means that no other company in the world can use your technology, right?
Alex: Correct.
Cristian: That’s amazing. In terms of the product market fit, do you think that you achieved it?
Alex: Very good question. Well, we started active marketing campaigns just a couple of months ago and till now our sales pipeline has already exceeded $150,000. So that definitely shows that there is interest. This hype about ChatGPT fuels businesses’ interest in leveraging AI, and so more companies are switching towards automating and streamlining customer support with the help of AI.
Cristian: Now I have an interesting question for both of you. I am curious to know about the biggest challenge that you, Roman, faced since starting the company and you, Alex, since joining the company. So what are the challenges that you faced? Of course, there are many challenges around the journey, but something in particular.
Roman: OK I can start. I guess that was when I mentioned previously that we were in the position of NLP Lab and not CoSupport. We were a laboratory for different researchers and investigations in the field of data science specifically in NLP. So, we were thinking, “OK very cool guys. We can do anything in NLP” and of course, like you know, “What does it mean to deliver the technology to the end user to the end line?” It’s like nothing difficult, but that was kind of the biggest challenge for my tech department because having kind of cool innovative, and state-of-the-art technology doesn’t mean that you can deliver quickly, and efficiently that technology to the end-user and that’s the biggest challenge.
Cristian: Yeah how about you, Alex?
Alex: Yeah from my side that’s for sure the demand creation. So, the market is not formed yet so it’s a challenge for AI companies that build AI solutions to educate their customers because there are so many companies on the market right now that could save hundreds of thousands of dollars by leveraging AI, but they just don’t know that so we really need to educate them. We need to prove to them that look, here is AI, here is what it can do, and here are the results it can bring to you, your company. So, to summarize demand creation is a problem.
How to Find the Very First Customers?
Cristian: Yes, I understand. How do you find your first customers?
Alex: Good question. So, when we started, we started with email outreach, but then there was an initiative from our director of marketing to start the outreach through LinkedIn. That was the first time we reached out to leads through Linkedin. How do we collect the list of leads to whom we reached out? We just went to different platforms where they compare different software and we found reviews left by different people on the AI-related companies. We created a list of like 50 companies whom we decided to reach out to because they were already using some AI solution and we reached out for the first time through LinkedIn and we actually got a couple of responses. One of those contacts converted into our client. They initially started with, “OK we don’t trust AI yet, but I like this solution that generates ready-to-replace suggestions for your support team.” So, we started with that, very small volumes and then they realized that the accuracy is 99.9% and they’re like, “Why should we double check if it works correctly 99% of the cases?” So we launched the automatic AI and they were like, “OK cool I think your search solution which is the BI will be as cool as the first two so let’s do that as well.” So, that’s eventually converted from a very small project to a paying client who is with us for a couple of months already.
Company’s Vision
Cristian: That’s amazing. Now before getting into some personal subjects. I want to know about the vision for the future of the company, and what you want to achieve in the next couple of years.
Roman: Well, for sure we want to leverage the latest technologies, up-to-date technologies, and state-of-the-art technologies so like in general all stuff, cutting edge, is for us. From my personal point of view, I think that we’ve made the first steps in the era of agents, AI agents and that’s pretty much important. What does it mean technically? It means that, believe me or not, in a few years we’re going to have AI agents, a kind of AI personal assistant for most progressive people, especially for businesses, especially for customer support. So, that’s really important and I think we’re going to have some kind of interaction between AI agent and AI agent or many AI agents. And really really, I mean many interesting things gonna happen, so that’s our vision.
Cristian: I think it’s super exciting. Especially with AI, you’re in a good position in the market.
Alex: Yes.
Cristian: What’s your favorite software that you use apart from your own?
Roman: Okay, the favorite software that we use. I mentioned that I began as a machine learning engineer as a data scientist, so from that time even now my favorite software is software from Jet Brains called White Charm. So, basically White Charm is an EDE software for creating compiling, and running software especially for Python. I just usually make a code review and so on so that’s my personal preference.
Cristian: Alex?
Alex: That’s a very difficult question for me to be honest.
Cristian: Any tool that you think, it can also be graphic design productivity.
Alex: I think honestly that Google Workspace is just top-notch. Like with all their products that I use on a daily basis not just email, Drive, Meets, Docs. Definitely Workspace.
Cristian: What would be your best piece of advice for someone just starting out at 20 years of age after finishing their studies and wanting to get into AI, into software, into tech? What would you advise them?
Roman: I guess that’s maybe a question for me because that’s also part of my backstory, how I became started in AI and maybe that’s gonna be kind of motivational for somebody else, I believe. So that started, I guess, in the third course of my university. After my lecture on Java programming, my professor discussed some opportunities with me, and said, “Hey, Roman, you’re a pretty smart guy and I know that our local IT company has data science camp science courses. So would you mind trying because you’re smart? You can after that get an interesting job in that field.” believe me or not, at that time I didn’t know anything about data science at all. I had high marks, I really loved programming, I loved computer networks, I loved administrating, and a lot of interesting stuff but I did not even like Python data science at all. After that, I tried I learn Python. I guess, within two weeks, I passed the test task, and I passed an interview. We had the courses after we had the science camp. Of 15 or 16 people, only two were left to continue moving with the company. I was in that pair. Basically that’s how I started my AI career at that time. That’s not the kind of AI for that time. The data science keyword was pretty much hyped not AI, but data science, and after that I just wanted to focus only on NLP because that was my choice.
Cristian: I guess those are my main questions for the podcast. If there’s anything else you want to share today before we close it, please share anything that you want to the audience. Alex: Yeah, thank you very much for having us today. I would like to add just one more piece of advice from my side. This one is for the companies who are looking for solutions. So, don’t be afraid to at least explore what AI is capable of right now because as I mentioned more and more companies will be switching towards AI and customer expectations will be growing. They will be expecting immediate responses and multilingual support 24/7. So, if you want to keep up with these continuously growing expectations, you need to start looking into AI. And believe me, AI can do a lot already. It can save you thousands of dollars already.
Cristian: Thank you so much for joining us and I wish you the best of luck with the product. I really believe in it and I hope a lot of businesses realize the potential of it, how to save time, make more money, and overall use a great solution with AI.
