Episode 1: Introducing the AI Agents Podcast

You can always tell ChatGPT to write up an email but you can't have ChatGPT actually send the email or book a vacation for you on TripAdvisor or whatever. That action portion is the key part about AI agents that really sets it apart, and even AI agents interacting together is another topic we can get into later in the podcast.

Hi, my name is Demitri Panichi. I'm a content creator, agency owner, and AI enthusiast. You're listening to the AI Agents podcast brought to you by Jotform and featuring our very own CEO and founder, Idin Tank. This is the show where artificial intelligence meets innovation, productivity, and the tools shaping the future of work. Enjoy the show.

Hey, I'm Idin, and this is the AI Agents podcast, a place where we're going to discuss AI, AI agents, and much more. How are you doing today, Idin?

Hey Demitri, I'm doing great. I'm really excited for the first episode. It's going to be great, I'm ready to go.

Just a little bit of background on the podcast. We've been discussing this for a little while. This podcast is brought to you by Jotform, and we're going to be discussing AI, AI agents, and everything in that realm. There are a lot of good opportunities for people to learn about these products and have them impact their businesses and lives. We wanted to bring this topic to the forefront.

We will be on YouTube, podcast platforms, and there will be short clips all across social media. We're really excited to discuss these topics with myself and Idin, as well as different founders and creators of AI applications and AI agents.

It's really exciting. How did you come up with the idea for the podcast?

I've been developing new software called AI Agents, Jotform AI Agents, for the last year. Because of that, I've been really interested in what's going on with AI agents, constantly reading about it and watching everything I can find. I gained a lot of experience while developing the product and building it with beta users.

I've been thinking about how to share all these things happening with AI agents and how to share this excitement with people because it's going to be big and getting bigger. I wanted to have a platform where we can share all the things coming out and the exciting stuff we see.

You and I did a podcast a while ago where I was talking about my book, Automate Your Bus, and about Jotform. That's how we met. Later on, we stayed connected, and you did some video projects for Jotform. When I was thinking about doing a podcast about AI agents, the first person who came to mind was you because you were also excited about it. That's how we started, the origin story.

I love AI, specifically AI agents. It's a great topic because it's popular now and I think the future. Nobody really likes interacting with AI in a way where you're actively typing things out but still doing most of the work. AI agents are the next level up in AI where it helps you take your ideas and have somebody else put them into action.

They're not quite people yet, but people are not used to using chat screens like ChatGPT to automate work. They can ask questions and get answers but don't do anything with that. AI agents give AI the ability to take actions on our behalf and get things done.

To clarify, you can tell ChatGPT to write an email but can't have it send the email or book a vacation. That action portion is the key part about AI agents that sets it apart. Even AI agents interacting together is another topic we can get into later.

It's a good turning point to discuss what inspired you to develop Jotform AI Agents specifically and what challenges you hope to address with that.

It wasn't actually the first idea. About two years ago, when I had my third child, I was away a bit from the busy work at Jotform. That allowed me to think about what's happening with AI and how we can use it within our product. I was really excited about that.

A year and a half ago, I took three engineers who had been doing AI for a long time and started working on the idea of helping people create forms with AI, basically a form builder co-pilot. It was going well, and we discovered people were okay with talking to AI and giving instructions on a chat screen.

We opened it to first 200 users, then maybe 1% of our users. People used the product a lot and got a lot of value. It clicked that we could help a lot of Jotform users. We have 25 million users and host many millions of forms.

Helping Jotform users is good but helping their users is even better. We decided to pilot this new project and focus on AI agents. The idea is that instead of filling out a form, you could talk to the form, talk to a chatbot or AI agent, and fill out the form by talking to it.

That's very powerful, especially on the phone where you can't fill out a digital or paper form but can talk to an AI agent over voice. We found the value is even greater if we allow people to train their AI agents with knowledge and give them actions to do things.

When we started building the product and getting beta users, people were really interested in instructing AI agents to do things the way they want. We planned to release the product earlier but saw people needed more, so we decided to launch in February 2025. We still have time before launch but have lots of beta users and are learning a lot.

I want to ask about the knowledge portion because it's key. You can create GPTs with knowledge bases and custom instructions. Why is it important to get down to specifics for use cases and actions? How do you do that in the product with Jotform?

It's the same way with Jotform AI Agents. You can enter your knowledge base by uploading files or entering text and instructions. What's different is the personality and visual features of Jotform Agents, like avatars and custom pictures, which make it stronger.

Forms are boring, so we wanted AI agents to come alive with personality and visual options. We also use RAG technology, which was a solution when including large knowledge bases in prompts broke down. RAG made our product much better.

One way we use Jotform AI Agents is to provide support to Jotform users with AI agents, but only in a small percentage because it's still experimental. Adding knowledge bases about Jotform's many features initially broke down but implementing RAG allowed uploading lots of documentation and improved performance.

Beta users use AI agents as chatbots for their websites, including e-commerce sites. The chatbot provides guidance to visitors and can fill out forms, take orders, register users, or send information directly. That's what makes Jotform AI Agents powerful.

To clarify, RAG stands for Retrieval Augmented Generation. It starts with data uploaded to the knowledge base and searches through it for semantically contextual information to add to the prompt context. This improves responses by providing specific information rather than generic knowledge.

Let's talk about the product itself. Idin will show it on screen, and I'll provide commentary for audio listeners. You can see the screen recording on the Jotform YouTube channel. The page is currently only available for beta users, but you can get on the waitlist.

We create a landing page internally for beta users to preview Jotform AI Agents. It's not just regular chat; it has personality with different kinds of avatars and templates. You can pick an agent or start from scratch. Agents can also fill out forms, not just ask questions.

We can create an agent from scratch or turn a form into an agent, which is useful for existing Jotform users. For example, we could create a feedback AI agent for our podcast to get listener feedback.

Creating an agent actually creates a form with form fields and some intelligence or knowledge you can edit. Here we have the AI Agent Builder where you can design, train, and publish your agent, including choosing an avatar that fits your case.

Customization is a core differentiation for Jotform. You can generate different avatars and customize your AI agent's persona, including name, role, language, tone, and instructions on how it should talk or act.

You can give the AI agent knowledge, like describing our podcast about AI agents and its features. People create AI agents and watch how users convert, then add knowledge like questions and answers to improve it.

People want to give instructions to AI agents, for example, if user sentiment is excited, ask them to review us on Apple Podcast. People role-play with AI, chat with it, and improve it by adding more actions and knowledge.

We have a listener feedback form created by the AI agent, which you can customize. The AI agent syncs with the form and can fill it out by asking questions. You can customize the form anytime and add more forms.

You can talk to your AI agent and give it instructions in a section called Teacher Agent. For example, if someone is excited about the podcast, ask them to review us on Apple Podcast. This is experimental but shows how actions are created.

Podcast Pete, the AI agent, asks how it can help with questions or feedback about the podcast. It tries to take action but initially wants you to fill out the form first. The form shows a percentage completed and asks questions in a conversational format, which is more palatable than manual forms.

Sometimes AI might have trouble understanding questions, so you can check the form to make sure questions are answered correctly. You can also fill out the form manually if you want.

You can edit the questions in the form. I answered them now: introduction to AI, five stars, and I loved it. The demo shows that chat is much more conversational and easier to fill out the form than traditional forms.

What could we improve for future episodes? We could improve by adding other people to the podcast, like AI agents news and founders. Would you recommend the podcast? Of course, we would.

What kinds of innovative use cases have you seen from beta users? Many replace their forms with conversational AI agents. Others put agents on their websites as chatbots to converse with visitors. Some use it in sales calls where junior salespeople get answers from AI agents during calls to help customers and fill out forms.

It's great to live in a time when if you don't know answers, you can cheat on a sales call with an AI agent. People also use AI agents for basic HR questions, like how company processes work, which saves time and resources.

For example, an HR app can have knowledge about company policies and forms. Employees can install the app on their phones and access the AI agent to ask questions or fill out forms. The app is a PWA that can send notifications and be installed from a website link.

On desktop, you can install the app as a shortcut icon that operates like an application. You can customize the icon and have your own AI assistant for customers or employees. It's very easy to use.

Many AI agent applications don't have standalone apps, so this is an upgrade. It allows specific questions and comfortable use, unlike general ChatGPT. It limits answers to your own knowledge base, avoiding random or hallucinated responses.

Our product philosophy is to release early and often to beta customers and learn from them. We watch how users use the product and define future plans based on their needs. People describe what they want the agent to do, and we add that to the product roadmap.

AI agents are the future. Just like web apps replaced installed software, AI agents will revolutionize software use. There will be many vertical AI agents for specific use cases like healthcare or teaching, and Jotform aims to be the platform for building custom AI agents with personality and sharing capabilities.

You've nailed the aesthetic and natural form integrations with knowledge bases and actions. Many products exist but none combine these features as well. We're excited to document this journey on the podcast and cover different products and topics with founders and CEOs of SaaS companies.

Thank you for joining the first episode of the AI Agents podcast. We appreciate you watching and listening, and we'll see you in the next one. See you next time, thanks.