165 | Making Data Collection in Forms Effortless with AI: A Deep Dive with Aytekin Tank of Jotform

0:00Hello, and welcome to Leveraging AI, the podcast that shares practical ethical ways to leverage to improve efficiency, grow your business, and advance your career. This is Isar Maitis, your host, and I've got a great topic for you today. We all, both in our businesses and our personal lives, have to collect information. We need to collect information from employees, from clients, from colleagues, from prospects, from friends when we're planning a road trip somewhere, from other parents in the school, from neighbors in the HOA, every time we go to the doctor's office, every time somebody opens a customer service ticket, all of these things are means of collecting information.

0:36But I don't know a single person, myself included, that wakes up in the morning and I'm like so excited that I need to fill up a bunch of forms today. That just doesn't happen. So forms have been our way to collect information and analyze information afterwards for a very long time. The problem with forms is that, first of all, they're static, meaning by definition, there's one size fits all for what you're trying to do, which means you're not collecting enough information on things you may want to do with specific individuals and you're connecting too much information on stuff in other cases which is not optimal for the people who are collecting the information whether again you're a person or a business and it's really annoying for the people who fill up the forms because you're like why do you need to know this information about me but it's a mandatory field that you have to fill out but all of this is changing literally right now with agents because agents can have a conversation with you on any channel.

1:29This could be text, WhatsApp, Slack, Teams, emails, and even voice. And that makes it A, a lot more fun, and B, because agents can understand what's going on, can be personalized for specific use cases for specific individuals based on just the information you need to do. So this is a game changer for all the use cases that I mentioned before, which, as I mentioned, is stuff that we are doing on one side of the equation or the other almost every single day. Now, as on this podcast, we focus on the how and not the what. Like, how do you, okay, this is a great idea, Zaya. I would love to hear more.

2:02How do you actually do this? How do you go from, I have a static form to I have an agent that can help me collect the right information in a better way. And there is no better person literally on the planet to teach us this than Itakin Tank. Itakin has been the founder and CEO of JotForms. So those of you don't know JotForms, JotForms is one of the leading forms platforms on the planet. It has been around since 2006, and it has more than 25 million users around the world. So they know one of two things about how forms work, and they're literally in the process of revolutionizing this whole concept of data collection with agents on the JotForm platform.

2:42And as I mentioned, that will change our interaction across so many aspects of what we do. A big one is definitely customer service where literally every company on the planet has some kind of a form to collect information about customer service. And all of this is going to change, or I'll be more specific, it's changing. It's not going to change. And so I think this is a huge part of the transformation that is ahead of us. And in this particular case, there's no downside. There's a lot of negative things in AI that will have impacts. But I think in this particular case, it's all for the better.

3:12It's better for the people who are collecting the information. It's better for the people who are providing the information. And it's better for analyzing the information afterwards, because it's going to be more specific. And so I'm really excited to welcome Aitakin to the show. Aitakin, welcome to Leveraging AI.

3:30In the next few years, AI technology will change our world dramatically. Whether you are a business executive trying to catapult your business forward, or just somebody who refuses to be left behind and want to advance your career, this is the show for you. I'm your host Isar Maitis, a serial entrepreneur and an AI enthusiast. You'll hear invaluable practical tips from innovative business leaders, AI practitioners, and some of the brightest AI minds in our world today on how you can leverage AI in ethical ways to advance your career and grow your business.

4:10Hello Isar, nice to be on this show. and yeah so it's really for the kind of boring and that's actually why i started job form like almost two decades ago i was working as a developer for a media company in new york and one of my duties was to create forms like i was creating all kinds of forms customer support forms contact forms questionnaires surveys all kinds of forms and i didn't like it i hated it and that's why i actually looked for a product that could actually take this away from me but I couldn't find one. I decided, okay, hey, this is the product that I'm going to build. And that's been 19 years ago.

4:49And I started JotForm. And today we are serving 25 million people, also over 50 million forms. So we have been building all these additional products and features for our users. And since the start of this new AI revolution, I was really excited. and I was like looking into like how we could use this. And so one thing I've seen, like AI actually gives you this kind of like, what makes AI so special is, especially these LLMs, these LLMs is actually they can speak and they don't just speak. They actually understand. They can understand. They can speak. They can really respond to you. They can provide information.

5:32So what's really happening here is that that now products and users can actually speak to each other. I think that's the big revolution. Everyone is focused on all these different topics like AGI or code generation with AI, things like that. But I think what's really special about this new revolution is you can now speak to products. And I think that's really powerful. And that's actually what I decided to do. In the beginning, I was thinking, okay, what if we could just talk to the forums? What if we could just fill out the forms by talking to them and just talk to a chatbot and then fill out the forms?

6:12That's how we started this new product, JotForm AI Agents. One and a half years ago, we started building this product. And that was actually pretty hard. I was expecting it to be easier, but it turned out it's just because we are entering data. We need to make sure it's correct. It turned out actually quite challenging, but we were able to actually do it. And we were able to create this product. and we started giving it to our beta users. But something happened that surprised us, that when we gave this to a small number of our users, we started 1 % of our users, what we saw was most people weren't actually using it for forms.

6:52There were some people who were using it for forms, but most people were actually using this as a customer service agent because this product also had the features so you could train your agent. So you create your AI agent, You can train it. You can add knowledge to it. You can upload files. You can actually make it use tools like appointments. But what happened was like most people actually start using this like a chatbot for their own organization. And they started like putting it on their website. And they were like just linking to it. And so most people were actually using it for customer service.

7:28That's what we discovered during their beta program. And even though right now, like the product is actually really good at filling out forms. So we are actually about to release this product in two weeks. And I don't know when this is going to be available, but it's going to be a product about the customer service. How can we provide customer service with AI agents? Imagine a world where every organization, every business actually has their own AI agents. and these AI agents can actually like talk to your users, your customers, your clients, your students, your patients, your, I don't know, like your volunteers if you're a nonprofit on your behalf.

8:09And it learns from you. And it actually takes action if someone needs to register because it already has, it can actually do forms as well. It can take registrations. It can answer questions about your, So you can even make purchases just talking to the AI agent. This has been like we've been working on it. We are really excited about it and it's going to come out very soon. Yeah, I'm really excited about this new product. Yeah, fantastic. First of all, we're going to dive into an actual demo so people see exactly how to set it up and what it can do. But before we dive into that, I want to touch on a few things that you said that are critical for people to understand.

8:46One is there are two things I highly agree with you or strongly agree with you as far as what are the two biggest things that I think AI will do, and it's going to do a lot of things. But from the day-to-day of our lives, what is it going to be the biggest difference? One is the fact that these tools know how to understand unstructured data, qualitative data, right? So previously, we all had the option to do quantitative data analysis, right? You could take numbers from somewhere, and if you were good enough in Excel or you had somebody good enough in Excel, you could analyze the data. Unstructured data, text, voice, we couldn't do anything with it unless you would hire McKinsey or Ernest & Young and they have an army of people in India and they will somehow try to make sense in 65 ,000 entries that you have in your ticketing system.

9:33Otherwise, it was just not doable. And now it's doable. AI is actually really good at analyzing unstructured data. So that's one aspect that I think it will completely change the way we collect and analyze data. But the second aspect is what you said is I think the concept of user interface is going to change from everything we know today, which is, okay, we have a keyboard, we have a mouse, we have a screen to something that is a lot more natural, like we communicate with anything. And I think that will be true for anything we know and it will take time. But why does my microwave need a keypad?

10:07Okay. I've put two pieces of pita bread in there, I want them defrosted. Boom. And it will do it. Like, why do I need to know what's the best time and what keys to press to defrost pita bread that has been frozen or pizza or whatever it is that you're doing? And I think that will change literally the interface of everything we know. Because the keyboard and mouse that we're so used to for the last, I don't know, 30 years were a necessity. Like we didn't have a different way to communicate with a computer, but that's not the case anymore. I've been building a software product that's been used by millions of people for the last two decades.

10:42And the biggest challenge has always been to make it easier to use. Because it's just, it's not people are not dumb. People are busy and people are overwhelmed. And they don't have time to learn your product. And if they could just talk to your product, it's just so much easier. So the speech is a new UI, in my opinion. Yeah, I agree with you. like because everybody knows how to talk like we even from the like the early age what we learn first is we learn how to eat maybe and we learn how to walk and then what do we learn my two-year-old girl is now starting to speak in sentences and just we learn and once we start learning speaking and listening to others understanding and communicating with others with speech it's it's pretty much everything we do it's it involves and that's what we know really best and And it's very second nature to us.

11:34It's very natural to us compared to trying to learn a product that every product you have to learn it. It's just so much, too much for people. Yeah, we agree 100%. I'll say one more thing specifically about data collection. Let's dive into the actual demo because I'm sure people are anxious to see how this is actually created. Data collection through speech actually gives you a lot more information than through a form. because usually let's take a entry form of a sales process. There's usually like a, how did you hear about us section? And it's usually a dropdown menu that has a word of mouth, social media.

12:10I saw an ad on LinkedIn, whatever. And the problem with that is, first of all, it's bias. I give you five options. Maybe you heard about me on somebody else's podcast, right? Like in our case, right? How do you hear about us? I heard you speaking on Isar's podcast, but that's not an option because you're not going to put 67 options there because the form will be not usable. The other thing is the nuances that you can get, right? When people speak, they will give you more information, more nuanced information than they will ever type in a form. And because AI know actually how to analyze that, it's not just, again, before you couldn't do anything with it because it was random text.

12:48But now you actually get a lot more information while making it easier for people to provide that information. So everybody wins. And that's why I'm really excited about this. But yeah, let's, I will let you share your magic and how it's actually done and how the product actually works. All right. Let me share my screen. So by the way, for those of you who are just listening, A, this is going to be available on our YouTube channel. So the Multiply AI YouTube channel, and there's a link to that in the show notes. And B, there is a, we're going to explain everything that's on the screen. So if you're not watching, don't worry about it.

13:19We'll walk you through what's actually happening. So you can keep on driving or walking your dog or doing the dishes or whatever it is that you're doing when you are listening to the podcast? I mean, there are some chatbots, AI agents, but when we were starting this, like one of the things that we really wanted to like really spend our time on was like give the AI agents a personality, give them these AI generative pictures. And like in the early days, like we were thinking about using robots and things like that, but we really wanted them to be like human, even though, you know, they're always starting a chat with the AI agent, like they're always telling the user that you are talking to an AI agent.

13:57We still wanted them to look like human. And we gave people all kinds of different options and they can generate, they can even turn their faces, they can use their faces or they can turn their faces. We just change it, apply different AI functions to use their faces as well. But we wanted to give this like a real friendly AI face. And one of the things that we are really investing in is having all kinds of different options for the users to be able to to be able to contact talk to the customer service agents you can talk on the phone or you can put it on your website like we support whatsapp texting sms we are working on zoom just we are going to be available it's going to be available when we release it in two weeks and we are doing all these different actions to be able to really use this AI agent in multiple channels.

14:52And I'm going to create an agent here, but we can also see it in action. There's one here, right? I can just start speaking to it. What is just from AI agents? I can ask it questions to find out more about it. And what we found was like a lot of people actually use it on their website to be able to answer questions from their customers. Like one of the most popular used AI agents was like one of the festivals, a famous festival actually used this JotForm AI agents and they were getting like so many questions about the festival, where people can park, what kind of different pricing options, like different kinds of questions about the live event.

15:35And we see that a lot, especially on the live event. Like people are like, they don't have time to actually like just read everything on the website. So instead, they prefer to talk to a chatbot and ask their questions and get answers from that. So let me start by creating my first agent. And here we are. We can create an agent from scratch. We can turn a form into an agent. This way, we already have the form ready and it knows about it. And this is especially useful for existing JotForm users. We also created all thousands of AI agent templates. So people can actually start from scratch. they can actually use user templates.

16:13Or this is one of the things that I actually use on my own website. I actually cloned myself. I created this chatbot. And basically, I put on all the knowledge about me, my articles, and so that people can actually talk here. And if they need to pass a note to me, if they need to just contact me for some reason, they can just write here and contact me. So this AI agent can talk on behalf of me. Yeah, so I think the biggest difference here, and for those of you, again, who are not watching, one of the options is clone myself. And I think the biggest difference here is in addition to company information, you give it your personal information, right?

16:53The way you write, your style, your tone, your history, all that stuff, other things you've written. So it now can speak and behave a lot more like you than a generic agent, right? That's the biggest difference here. So, yeah. Do you have a question? No. Okay. Let's build one. So let's start from scratch. Let's think about what kind of agent we can create. Maybe we need to create a customer support agent for our website, or maybe we need to take an appointment. We could go out with this customer support agent and just we can write a prompt. This could be about like we can describe like our purpose, like what kind of customer support agents we are going to have.

17:32We could just start with this is for my small business. I do vacation. I don't know, rentals, right? So we could actually provide more information. We will have an opportunity to enter more knowledge, but we could just provide all this information here. And then it will go ahead and it will actually, it will also create forms if it sees that a form would be applicable, but it doesn't have to be. And so it created my agent. This is our AI agent builder. This builder actually has, you can actually customize your agent with different personalities. You can generate images for your agent and you can actually change the style, change the colors.

18:18Whatever fits your case best, you can actually use those colors and style. And so this is the first screen and we can actually preview what our agent looks like here. So here we are, we have this customer support AI agent. And it's just because we don't have too much knowledge, it just starts very, very generically. But we could answer knowledge about our business. We could add all the questions and answers about our business. So we can actually answer those questions. Let's dive into this for a minute, because I think that's the biggest gap between what historical website chats were and what they are right now.

18:55So historically, you had to build this decision tree of what the chat will do at any given step and what questions it has answers for because the idea of a website chat is not new. When I was running Last Minute Travel a decade ago, we had chat on our website. But it was a pain in the behind, like clearly, because it only knew to answer stuff that it knew more or less word for word what the question is going to be. and then it had an answer and then there was like a decision tree of which other things to do and whether to refer to something else or what links to give them and stuff like that so let's let's dive for a second into how you quote-unquote train the model in your platform so people have a better understanding of what it is now and how it is different than what it was before yeah so that's the train tab and in the train tab we could give ai agent a persona we could give it a name different name and just different options we can prompt it to be like shorter give short answers or really minimal answers or long answers and yeah there's also i can see it you can define whether it's friendly or professional or funny like you can give it an actual personality yeah since this is this is going to be about like a vacation rental customer service agent.

20:23We could actually enter knowledge base about that. And let's say question and answers. Maybe we provide what kinds of rentals you have. Maybe rentals you have. I got to answer a question about the Q &A section. If I don't fill it up, because now I can add questions and answers, but let's say I don't fill it up, but I do give it all that information in PDF files or my website and so on, it would still know how to answer these questions, right? That's the whole point. Yeah. I could just, we could just enter a URL and it could actually read from that URL. Yeah. So I think that's the biggest difference, right?

20:58So the biggest difference when you prep your AI is that you don't need to give it Q &A unless you have very specific questions for which you want very specific answers. Like what is your cancellation policy? You don't want the AI to figure it out. You want it to be very specific because that is going to cost you money if it gets wrong. But I think knowing about your business in general is as easy as, okay, here's my website. Here is our terms and conditions. Here is our PDFs explaining about, in this particular case, each and every one of our properties, as far as how many rooms it has, how many beds does it have, how many people can sleep there.

21:37What's the time you can check in and check out all these things. Usually, if you're a rental company, then you would have a PDF for each and every one of those, or a web page for each and every one of those. And I assume it can just learn from that, and then you don't need to write all those questions and answers. Exactly. And that's like the very common case. People will just enter their URLs, and from there, our AI agents will actually learn their whole business, and it will be able to answer questions. And again, people are busy. They don't have time to read, find their answers in a very big website.

22:12So instead of doing that, they can talk to the AI agent. They can call up the phone number that's shown on the website and then talk to the agent on the phone and get their answers in a similar way. Cool. So let's, yeah, really let's dive into actions because I think, again, this is where traditional forms were very limited or chats were very limited. And I think now it can actually do a lot more. Yeah, here we actually, we have all these, basically you can actually build, give it more instructions, right? You can actually instruct your AI agents, take some actions. For example, when the conversation starts, what can we do?

22:50Like we could show a button, like we could say an exact message, say something like, welcome to our website. How can we help you today? Yeah. So basically this is like a very broad, if this, then that, that allows you to define, but what other actions that are beyond the chat can you take? Because I think one of the biggest differences between a, again, traditional chats and agents is that agents can use other tools. They can actually go beyond, yes, go and I know what to answer in this case, or I know what to, who to refer you to. It can actually do more stuff. Yeah. And then we can actually test this We can just go to the preview page to see if this works.

23:37Welcome to our website. How can I help you today? Right? It just, it folds our instructions. And I want to find out about your rentals.

23:52It doesn't have too much knowledge, but it remembers that we have two SKU rentals. We could provide more information and can provide details to the users. basically it's about training your ai agents and you know what i have an interesting question for you about feeding it live data so let's use your example of two ski rentals will it be able to connect to like my reservation system and see if there's what's the availability in february for midweek uh days will it know how to answer questions like that to actually go and search a live database from somewhere and then provide answers based on that that's actually how we what we are able to do with using tools.

24:32So it can actually integrate with Google Calendar, and it's able to integrate with your existing Google Calendar so that you can actually, similar to Calendly, it can, if you need to schedule. And it doesn't have to, if you think about it, like if you are using, I guess you could create your, for your vacation rentals, you could create different Google Calendars to keep track of them. Or Jotform also has its own like appointments fields it's without the there's actually it's actually possible to use this okay this is the one right it's possible to do appointments even without the google calendar integration basically you can actually set up your appointments i guess we'll need to set up custom minutes right we need to set this up so that it's actually instead of hours it we need to be able to like book days, right?

25:22Yeah, yeah, yeah. Make it relevant to the specific space. Okay. Yeah. It's something, right? Yeah. But let's go back to the tools for a second, because I think that would be interesting for people to understand because one of the biggest differences again about AI is that they can use all these tools. So in this particular case, it can take note, it can set an appointment, it can send an email, it can trigger a workflow, which is some kind of an automation. It can send messages on Slack and Teams and stuff like that. It can show a video. It can find a website. It can find information in a website.

25:54It can show lists. It can send API requests. So again, if you have a platform that is not integrated into this yet, you can use an API. So if you have, again, your own existing ticketing system or your own scheduling system or your own whatever it is that you have, ERP, like it doesn't matter if somebody wants to ask you about an inventory, does it exist right now or not? This can query through an API or ERP and provide an accurate answer. So it's one of the biggest differences of agents versus just traditional chats is their ability to A, take actions and B, use third party tools because they understand the request, right?