DRUID Talks Ep #1 Busting Myths about Conversational AI with Daniel Balaceanu
DRUID Talks Podcast episode 1 touches on the misinterpretations and misconceptions related to conversational AI.
Episode #1 of the DRUID Talks webcasts features Daniel Balaceanu, DRUID's Chief Product Officer, and Subject Matter Expert, Kieran Gilmurray. See the full transcript and episode below.
Kieran Gilmurray: Hi, this is Kieran Gilmurray, and you're watching DRUID Talks Webcast Ep# 1. Our first guest is Dan, the Chief Product Officer of DRUID AI, and today's topic concerns the misinterpretations and misconceptions related to conversational AI. In many mistaken ways, conversational AI is limited to being considered a chatbot or a chatbot technology when it’s actually used for a wide variety of purposes, and that's business cases, and incorporates anything and everything AI-related and not just simply NLP. We're going to go to the root of this technology to try and uncover questions like what is conversational AI and how it is more than a chatbot. Is conversational AI capable of being creative? Is conversational AI replacing human conversations or stealing jobs? And then delving a little bit later into the future of conversational AI in general.
So, Dan, we will not be bored today, shall we say?
Dan Balaceanu: Definitely not. Definitely not.
Kieran Gilmurray: Well, let's jump straight in here, and I'll jump in and ask you the first question if that's okay with you.
Dan Balaceanu: Please. Yeah. Happy to be here to start with.
Kieran Gilmurray: Yeah. It's good to see you here. Actually, our first episode ever, and you're in the driving seat, so this should be a bit of fun for everyone listening in as well. Dan, if I ask you just an initial question - and that's we're going to bust some myths around conversational AI today - would you, just for those people who haven't seen conversational AI to date, could you tell us what is conversational AI? What's it composed of, and how does it actually work?
Dan Balaceanu: Yeah, I mean, I'm almost sure that each of us, we already interact with a virtual assistant, with a chatbot or with a voice assistant over the phone. So, I think it's everywhere now. We just look inside and understand what is there. And within the conversational AI package, actually, there is a set of technologies that all combined deliver a solution, and many times, the conversation is understood like language, natural processing, the understanding of natural language. And this is true. This is the first part inside conversational AI: to have the machine understand what humans are saying, are asking. But then, once they understand, and this is natural language processing or natural language understanding, they need to respond. And this is the dialogue management, the dialogue flow and the capability of the machine to keep a conversation.
And then, in order to respond, they need to take the information from somewhere, either to build the information, and we will see the generative models and how visual assistants create content or from where to take the information and give it to you.
So, the main components of a conversational AI platform, and we will understand later, of a conversational business application, it's language understanding - to understand what humans are asking; it's dialogue management - to keep a conversation in the context; and it's generating and delivering the response and integration with the back-end systems because we are talking about conversational AI within the enterprise workspace, not in general.
Kieran Gilmurray: I really like that explanation, and I love the piece, Dan, as well, when you mentioned about retaining that conversation because agents today get a lot of different questions, and some may phone in the morning and may phone tomorrow, and may phone the next day, and they may end up talking to different agents in the contact centre. That's just the reality. And therefore, to retain the essence and the content of that conversation across different communication times and methods is really essential today.
Dan Balaceanu: And it's very important in the enterprise because it's not just the conversation “what’s the weather today?” It's about data, it's about processes, it's about applications, and almost all the time, the answer is related to the human asking the question. So, it's so important the context and the integration with the back-end systems.
Kieran Gilmurray: I like that! You mentioned a moment ago, Dan, about conversational agents and that they're not chatbots. Many times conversational AI can be considered a chatbot or mistaken as chatbot technology when realistically, they're quite different things, with conversational AI having a lot of different business applications. So, would you mind giving us a little bit of detail about the distinction and helping people understand where conversational AI can be used?
Dan Balaceanu: Yeah, I mean, there are multiple angles to look at the topic in front of us, but just from the terminology perspective, chatbot means engaging with the machine through the conversation, through the chat, but only from that perspective is not sufficient because right now you can call a bot and you can have a voicebot and have a voice conversation. And, as in real life, communication is done not exclusively by speaking, it's also the mimics. It's part of the conversation. Sometimes I'm drawing a picture to highlight and deliver more information, so it's not just chat and simple conversation. It's different forms of conversations, and this is a multimodal experience. Sometimes I am asking the bot, “Give me all the information you have about customer Daniel” and the bot says, “Look, this is the picture with all the information you requested,” and it’s building a screen for me with all the data, it's moving my experience in the digital workspace where I have tens of fields on the screen, I have buttons, I can click, I can interact with the bot, with the virtual assistant in so many multiple ways.
And there's not only that. I see information, I can change the information, I can trigger processes, and these actions are received by the bot and are further pushed into back-end systems. So, I'm completing business transactions. So, from this perspective, you see, conversational AI moves to conversational business applications and an end-to-end solution where the voice or the chat is just one of the actions I can use to interact. But the huge power is what the virtual assistant can do for me.
Kieran Gilmurray: That's really a key difference because chatbots, to some extent, when they've been badly designed - and we have to be clear on that point - when they've been badly designed, have led people not to feel positive about chatbots. Whereas, what you're describing there is putting in layers of intelligence in the enterprise, linking back-end and front-end applications to allow you to have an intelligent conversation that feels very humanlike, or that's what you're describing, Dan?
Dan Balaceanu: Exactly. And I mean, I'm totally open. We have engagements where the audience says, “Yeah, but look, I'm not happy about chatbots. I heard about them, I don't like them.” And then, when they see how a digital assistant is delivering a business solution, they say, “But this is not a chatbot, this is a business application, and they are right.”
And the issue is that many years ago, because chatbot technology is not something new, it's been on the market for many years… but at the beginning, the job was just delivering simple FAQs: what's the weather, what they can do, things like this. But this is far less than what is required. And you cannot have a business conversation if the virtual assistant is not capable to understand my context, to perform actions on my behalf. And this requires not only language understanding, it requires integrations, it requires automations, requires user experience. So, it's the full combination that makes the experience great.
Kieran Gilmurray: What you're describing there, Dan, almost sounds humanesque. Some people may feel that conversational AI and conversational business applications are designed to replace real conversations and, therefore, real jobs. Is that the case? Can conversational AI be more creative than humans, or is that belief system not accurate or not correct?
Dan Balaceanu: I think there are more questions. But, to look at the first question, “Is the conversational AI stealing human jobs?” Totally not. It's the other way around. Conversational AI, as other technologies, is empowering humans; are allowing humans to do the tasks faster, quicker, and better. Me as a customer, I can schedule an appointment with the doctor, I can sign up for a new contract just by discussing with the machine. Me, as a human agent in the contact centre, I would not be replaced by a robot. I will have a robot helping me to better serve my customers because it gives me quick information. It can take actions based on my instructions. So totally, robots and conversational AI is not replacing humans, it's empowering humans.
Now, what virtual assistants can do - if they can build content and create content? Maybe we have some time to discuss about ChatGPT and all this, the new momentum which proves that yes, technology and virtual assistants, conversational AI can be creative and create content. So, they create poetry, they can create a picture, and they can write whitepapers.
However, now we are discussing about conversational AI within the enterprise workspace, where the data should be 100% accurate and traceable. We don't want, in most cases, a black box that delivers an answer without explaining how that information was produced, from what source and have the entire traceability. So, in this regard, I see more appreciated a virtual assistant that can grab the right data from the right source and deliver it to the conversation rather than create the data on the fly.
Kieran Gilmurray: And I think that is a key difference because there has been a lot about ChatGPT and we've all tried it and it is - when we use the word creative - it's been fed a lot of existing data and then it's reassembling that in a different way. But that's a very different use case as you're describing, where the data is already in the enterprise. It has to be accurate; it has to be traceable; it has to be secure, and businesses have to understand what it is they're using to make real human impacting decisions.
Dan Balaceanu: Yeah, and don't get me wrong, I am absolutely amazed about the momentum and what ChatGPT is able to deliver. And from my perspective, I took three or four lessons learned from this momentum.
First, it's an amazing validation of people willing to adopt new technologies. I mean, look, in a couple of minutes, five million people signed up and chat with ChatGPT. 5 million people! Now it’s no longer a prediction, a forecast that people will prefer to interact with the machine. It's happening right now, and this is happening in the business as well. Employees, and customers are just waiting to have the opportunity to engage with a virtual assistant that can do things for them. And the need is there. We just need to respond.
Second, these big models are capable of understanding the intent without being trained. So, I tested hundreds of intents from business, and all of them have been recognized by ChatGPT. However, when the response was generated, obviously, it was generated with the generic information grabbed from the Internet because that was the data source. Definitely, it was not the expected answer within the enterprise. But intent recognition was fantastic.
So, in DRUID, from a technical perspective, we will integrate this forward but put it in the context of the enterprise and just use the part which is valuable for the enterprise and complete with all the other work which is already done in the DRUID platform. So, my view and I shared it, and it was the vision within DRUID from the beginning, there is no single technology to deliver the full stuff. The perfect way is to have an AI and digital hub and a virtual assistant capable to absorb and combine in a perfect way all the current innovations today and remain open for the future and give the end users and the businesses the flexibility to combine technologies.
Kieran Gilmurray: I love the open nature of that, Dan, and I also love that for those folks or employers sometimes, who worry about virtual assistants or conversational agents and will people use them or not - I think that was the fastest sign-up to any technology that we have seen in the last number of years, which really shows that people do want to engage with conversational technology and digital assistants in their daily lives as well. Exciting times!
Dan, as you mentioned a moment ago, conversational AI has been around for a number of years. It's constantly evolving and constantly improving. What does the future of conversational AI look like?
Dan Balaceanu: First, it’s not an end of the road, so conversational AI continues to evolve, and there will be innovations and improvements from different angles: language, understanding, voice understanding, speech, synthesizers, automations... And so, the whole industry is moving forward.
How I see it applied, I see a transformation of business applications. Conversational AI is not used just as a set of technology - language understanding - it's used in a combined way to deliver an end-to-end solution. And I see business applications transforming. We see conversational customer relationship management, we see conversational human resource management, we see conversation ticket management, conversational contract onboarding. It’s the traditional applications but built now in different ways.
And I see conversational AI everywhere. Look now, if I take a phone which doesn't have a touchscreen, I consider that this is an incomplete phone, a broken phone. The same if I were a customer visiting the website of a company and I don't have a chat, I don't have the possibility to speak with that application, I consider it incomplete. So, conversational AI will be everywhere in each system, in each business.
Kieran Gilmurray: I love the sound of that! I think the conversation is the most natural thing that we, as humans, enjoy the most. So, I look forward to dealing with more conversational agents as well because I want service 24/7. I want it on my phone, I want it on my computer. And I love what you're describing today, Dan, because the conversational AI components, the NLP, the NLU, the conversational business applications are as far away from a chatbot as I have seen in many, many years in this industry.
I wish you, and I wish DRUID every success. Exciting times for the company and exciting times for this industry as well! And conversational AI is definitely something that folks should look seriously at if they haven't investigated it to date. Thank you, Dan.
Dan Balaceanu: Thank you very much, Kieran. I'm also very excited, and I'm passionately living each project, learning from each project. And I am happy with our customers and partners, seeing what can be achieved right now.
Kieran Gilmurray: Thank you, Dan. Thank you, everyone else, for listening in to DRUID Talks, the first episode. Today we covered all the misconceptions related to conversational AI and the many mistaken thoughts around conversational AI just being considered a chatbot technology when in fact, it can be used with multiple applications in a wide range of business uses. So, today we covered why conversational AI is more than a chatbot; we talked about ChatGPT, is conversational AI capable of being creative, is conversational AI replacing human conversation or stealing human jobs. And we were very clear about the fact that it's not. It's actually augmenting great agents in context centres or in a wide variety of business applications and allows them to do a better job. We talked about the future of conversational AI and it is looking exciting.
And for those businesses who haven't thought of it, we also mentioned that this is a go-to hot technology that you should be considering.
Episode one is now done. A brilliant talk with Dan. Dan, thank you so much indeed. More episodes to come from DRUID talks over the next number of weeks and months. Folks, please check this episode out. Please check out future episodes. You will learn a great deal about conversational AI, conversational business applications and technology in general.
Have a perfect day!