Skip the Waitlist: Start using these AI solutions in your Contact Center
Colm: Hello. Good afternoon and welcome. I'm Colm Ó Searcóid, the content marketing manager at LeadDesk and I want to welcome you all to a sneak preview of features that help serve customers and improve efficiency at your contact center. Presenting today, we have Miikka Haavisto, Director of Business Development at our AI unit. Miikka has previously worked at Microsoft and Nokia and has extensive experience as co-founder and COO of an AI chatbot company. Welcome Miikka. Miikka: Thanks Colm and welcome everybody to today's sneak preview.
As Colm mentioned let's start by going through the agenda for today. For the ones that are not very familiar with LeadDesk, we will be giving a quick introduction to the company first and then dive into what we have been working on in the AI space for several years. Next, then we will cover our vision for AI development, where we see how it will impact the contact center space going forward. And after that is the most interesting part, live product demos. At the end of the session, we will have a possibility to ask questions or
actually you can ask the questions in the Q&A functionality throughout the presentation and then we will answer those at the end of the session. Hopefully there will be plenty of questions and if there are too many to answer now, then we will come back to those offline afterwards. All right. So a few few words about LeadDesk. LeadDesk is the leading European Contact Center company serving Enterprises and SMEs alike in Europe, for about 14 years now. While we do have a heavy focus on the European market, we have customers from Australia to US and Latin America as well, but really focusing on the European market.
We have plenty of happy customers and operate all over the world, and we are super proud for our safety certification. Safety and security is very deeply rooted in LeadDesk and I'm proud to say that we are both ISO 27001 and SOC 2 certified. To put it simply, we build smart software for better Customer Service and Outbound Sales. This next slide is one of my favourites from the LeadDesk's catalog of slides.
I call it the 360 View. And I think it really illustrates the space that LeadDesk is working in. We offer a fully-featured Omnichannel Contact Center solution for Customer service and Sales. Our customers can communicate with their customers in every channel from voice, SMS, email, web, chat, and other messaging services such as Facebook Messenger. You can easily integrate LeadDesk as part of your business processes and workflows with our very well documented APIs. And then on the bottom part of the page about workforce engagement management, those are tools that help motivate and increase agent satisfaction in the contact center workforce. That's a very brief
introduction to the company. And then we will jump to the topic of today. So what have we done so far with AI? The rise of GPT and large language models. So LLMs started an AI hype during the beginning of this year or end of last year that we have never seen before, and for a really good reason.
We at Lead Desk have, however, not been waiting for such groundbreaking technology to get started, but have been working with AI Solutions already for multiple years. In the next couple of slides, I will go through just two examples of the solutions that we have been working on. So first Chatbots.
And as Colm mentioned, I have a background in this space, so they are near and dear to my heart. LeadDesk has been or is one of the first providers of chatbot services in the Nordics and I would say also in Europe. We have been helping customers improve their customer service operations productivity already from 2016. Our chatbots are used in a variety of industries, including insurance, healthcare, accommodation and public sector. We have quite a quite a wide scale of public sector customers, especially in Finland. A nice little detail on the volumes.
I asked our team to check how many conversations our solution handled during 2022 and our customers who are building their chatbots using our chatbot studio handled over 2.5 million conversations in 2022. And to make it even more interesting, many of our customers, both small and our largest customers, reach up to 80, even, even 90% of automation levels with the service. So really nice, nice numbers there. Of course, I'm assuming that many of you on the call are familiar with chatbots, but just a really brief introduction: Chatbots help, first of all, by broadening service hours. So you can offer 24/7/365 service, which is not so easy to do with the European labor costs. There are a couple of different ways you can implement them. You can do a standalone service or then have the chatbots work together with the customer service agent, like for example, in Lead Desk.
Once the chatbot has done its job, potentially just gathering details and trying to identify the customer problem, if it's something the chatbot can't solve, it can be easily transferred to the agent. And this is probably quite familiar to all of you. The next AI solution that we have been already in our product portfolio for many years is the Predictive Dialer system. We have offered this already since 2019. How it works is that there's an algorithm that uses a multitude of different data points for conversation length, for example, and evaluating when certain agents might be available to take the next call. This is very efficient for high volume outbound sales processes.
It means that your sales agents spend a little less time in listening to dial tones, and more time talking with your customers or prospects. Predictive Dialer helps sales organisations to maximise the efficiency and output of their organisation. Then about LeadDesk's vision for AI development next. We could summarise the vision in this text.
So we want to "Provide intuitive AI-powered solutions which delight customers and make customer service and sales agent agents superstars of their profession." I think this summaries quite well how we see AI impacting the contact center world. And if we dig a bit deeper how it will impact, we believe that it impacts basically everybody involved in the in a contact center, whether it's visitors who are communicating with companies via contact center solutions.
If it's agents who are working in the contact center or the managers who are trying to optimise and follow the performance of the team. So if we start from the visitor point of view, with the help of these new technologies, we believe that visitors are able to get even more personalised and accurate support with the help of the automation. What is also now different compared to maybe a couple of years ago is that we believe that customers can get this automation service in any channel, including voice.
So previously automation has been maybe limited a little bit more to chat. There has been some companies offering email automation as well. And of course some voice automation. But we believe that in the short term there will be voice bots increasingly available on the market as well. What is also very beneficial for the visitors is that they can get service in any language. So not only in the language that most contact centers are capable of serving
with, let's say English, but you can use your own language and you will get help in your own language. Then on the agent side. We definitely do not believe that these new AI technologies will be the end of human agents in contact centers. What we do believe very strongly is that new AI technologies will improve agent efficiency working in contact centers, so will help them in many ways to complete tasks quickly and efficiently based on the available data, historical conversations with the customers and other data available about the customers, for example, in CRM Solutions.
Agents as well can answer the questions in any language. So they use their own language. It will be converted then to the visitor language. They will be benefiting from AI drafting responses for them, which will then increase the speed of serving the visitors. And, we can't forget the managers who are trying to make the most out of their resources and serve their customers in the best possible way.
This is just a small glimpse of how we see that managers will be benefiting from this. But we believe that with the help of AI, we can provide more insightful analytics for managers, for example, AI-assisted sentiment analysis. Then we have about half of the time ready for the demos and this is the most exciting part. So let's dive right into it. We plan to show a couple of different use cases, so I will try to be mindful of time to leave time for questions.
And if you feel after this, that I was going too fast in some things just let us know and we can arrange a separate session to go through some of these in more detail. We will cover three different areas. So first around the AI Chatbot, I will be sharing a solution where we build chatbot content automatically and then a live use case where we use the large language models or GPT models to rephrase customer questions in a situation where chatbot doesn't really quite get the correct answer. Then email auto creation. This is very exciting.
And finally, VoiceBot, which is a prerecorded demo today. And just one second, I will share my browser. Could you Colm confirm that you are able to see this kind of a list view? Colm: Yes, I can.
Miikka: All right, great. Thanks. So, one of the first implementations of GPT that we thought of at LeadDesk is that how could we utilise the powerful language models without jeopardising the accuracy on the end -customer side? If you have been following the discussion around large language models, you have probably come across a pretty big problem in my opinion, that the model tends to hallucinate a little bit. So what that means is that it will very confidently give you an answer, even though the answer might be incorrect. And the reason for that is that it really doesn't care if the answer is correct or not, or it doesn't even know.
It plays with statistics, and it provides the best answer it can provide. So how we tackle this is that we will harness the technology to create the chatbot content, version one. And then there will be a human validation round.
So agents will go through the data, make sure that the answers are correct. And then once that validation is done, the data will be uploaded to the chatbot. And there you have your ready chatbot, ready to be taken into use, of course after configuring the user experience a little. So what you can see here is an example of this UI. So basically what you can do is add files, so word documents, PDF files, or for example, add links to your website. If I just quickly give an example here.
You upload a document with multiple links. You decide how many questions you want per URL. You can decide how many alternative example questions you want per URL. So, sorry, Questions and Answers. And then how many alternative questions.
And then what is quite interesting and important for chatbots is the length of the answer. Normally about 300 characters is a good number for an answer. So this data here is taught with LeadDesk website data, and you can see that there are some yellow bubbles and green bubbles and gray bubbles. And basically what this means is that these have been checked by a human and they have been voted through, and that's why it looks like green.
This has been done halfway and this one has not been started. But if we, for example now, then select one example here. So AI Chatbot. So these are questions and answers created with the help of
GPT models using our custom prompts and only using website data. Here we have limited the number of questions to two, But you can see already here that it gives accurate answers. If they would not be accurate, the bot trainers would then go through this and edit. They can edit both the question and they can edit the the answer if the answer needs some editing. Once they are done, they can upload the data to the Chatbot Studio.
And this speeds up the deployment process of a chatbot from, potentially a couple of months, to just a couple of days. So it's a very huge impact in the duration of these projects. Next, what I want to show is the Rephrasing of Questions. Many of you have probably talked with chatbots. So this is nothing different or special. So chatbots can answer questions even with some minor typos there.
So but what happens when a chatbot gets a question that it has not been taught and it's not close enough to the content that has been taught. So let's try, this is always interesting because the models never react the same way. So I know Colm cringes a little bit because he doesn't like this sentence, but you may have seen that there was a small delay there.
And what that delay basically did is it was when the chatbot didn't initially find an answer. But what happens here is we send the question to our own GPT model in EU, and ask for alternative ways of asking the same question. And what we can see here in the chatbot training studio, unfortunately don't have a lot of time to go through but can show a neat feature here, nevertheless . Here you can see the question that I just asked or the conversation that we just had. You can see that there's a question, "is there a purpose to living?" And it has been flagged as an unanswered message. But then this alternative way of asking it came from the GPT model, which then we found the correct answer, which is "According to The Hitchhiker's Guide to the Galaxy by Douglas Adams. The meaning of life, of course, is 42." Another neat feature here, I can teach this sentence that it didn't know directly to this intent.
So this is an existing feature that we have had already for a long time. But this is a very powerful feature to continuously teach the chatbot. All right, now I'm a little bit short on time. So hopefully you were able to follow that brief racing demo there.
So in short, if the chatbot doesn't know the answer, we try to find alternative ways of asking the same question and trying to find the answer then. Very powerful stuff. The next one is about email automation.
So you can see here that there's a message that has been received, email that has been received from Laura. Laura needs to pause her account for summer because she's going travelling. This is a normal use case in sports or gym chains. Hopefully you can see in the screen or on the screen that there's this blue pulsating square around the "Type Your Message" box. What this means is that the AI in the background is telling me that there is a potential AI generated answer available for this email. I will quickly use a hotkey.
You can see that it's generating, and voila, there we go. We get the answer from the AI model. If you want to tune it a little bit, you're free to do so. Maybe type 'have a great summer,' etcetera.
And then you can send it out. I'm quite excited about this one. It is a very powerful, powerful stuff.
Our model has now been taught with a few examples and it works really well, and we are very excited to get this to our customer's hands first. Some customers are already starting to use this this week. So this is a very good example of of this intuitive way of helping in the in the contact center space. Next, we will go back to the Powerpoint.
And show the Voicebot demo. I'll just find the correct slide. Here we go. Hopefully you can see it. And as mentioned, this is a prerecorded demo this time.
You might recognise the voice in this one. VoiceBot: Hello. Welcome to Flight Booking. How may I help you? Colm: Hey, I'd like to book a flight to Stockholm please. VoiceBot: Great. When do you plan to travel?
Colm: I'll be going out tomorrow and coming back the following Friday. VoiceBot: Perfect. Let me check the available flights for you. Can you please provide me with your preferred departure time? Colm: I'd like to go out some time in the morning and maybe come back in the afternoon.
VoiceBot: All right. I have found a few options for you. There is a flight departing at 9:15 a.m. and arriving at 12:30 p.m.. The return flight is departing at 2:45 p.m. and arriving at 6 p.m..
Would you like to proceed with this itinerary? Colm: Yes, please. VoiceBot: Great to finalize your booking, I will need some personal information. Please provide me with your full name and date of birth. Colm: Yeah. My name is James Hargreaves, and my date of birth is the 21st of June 1990. VoiceBot: Thank you for providing the information. Your flight is now booked.
The total cost is $350. We will send you an email confirmation shortly. Is there anything else I can assist you with? Colm: No. Thank you so much. VoiceBot: You're welcome. Have a pleasant journey.
Miikka: Great. That was a prerecorded VoiceBot demo. And if you didn't recognise the voice, it was, of course, our host, Colm. So we have a few minutes left and Colm, do we have questions in the QA section. I'm just curious, how much time do I have? Colm: You have about three minutes.
And I do have one or two questions. Would you like to go over them now, or after this slide? Miikka: Then I will then spend maybe just one minute on this one. So just very briefly covering what else is coming from LeadDesk. So this was already mentioned in our vision. So multilingual chatbots and Live Chat is something that we and our customers and prospects are super excited about.
Hopefully we can share some more info about that later on this year. And then what you saw in email, this auto generated responses, the same features will be then available also for Live Chat later on. From the reporting side, we already have pilot customers using what we call QueryAI, which means that you can use natural language to tell the system what kind of reports you want to create, and it will assist you in creating those reports.
So that's a very nice feature for the admin side, or manager side. And then finally, on gamification, we have a feature where, if a customer service agent is, for example, handling a call, you can guide them to cover different topics and by identifying keywords in the live conversation there will be traffic lights that will turn green during the conversation. So it's a nice gamification aspect, but it also makes it simple for the admins and managers to follow that topics agreed to be discussed are actually discussed. So that's a very nice feature as well. That's that's it from my side.
Colm If you could please ask the questions. Colm: Yeah, I think one of the questions is a concern about data. So, what happens to my company data if I put it into this AI? Miikka: That's an excellent question.
Like I mentioned at some point in the presentation, we have our own instance of GPT model in our own cloud in Europe. So data aspects have been taken into consideration. So the company is providing the models, they don't have any access to the data. Customer data is not used to teach any models. The customer data is simply just used to help the customers automate their processes, create their chatbots, draft their emails.
Colm: Okay. Thank you, Miikka. And then another one is how do we get started using some of these solutions? Miikka: The easiest way is that if you're already a current LeadDesk customer, please contact your account manager and they will then start the process of helping you out. The alternative way is to ping me and I will make sure that you will get help immediately, so everybody can just contact me directly and I will make sure that your questions are answered. Colm: Thanks, Miikka. It looks like we're out of time. Just about. Thank you, everybody, for attending.
I hope you found this very interesting. And feel free to reach out to Miikka or myself or anybody at LeadDesk if you have any questions. Thank you.