Good Neighbors Episode 1:
From AI Experiments to Real Business Results: Jform + OpenAI

AI is transforming the way businesses work. But where does the real value begin? In this episode, OpenAI's Thomas Kubeck joins us to discuss how OpenAI and Jform are turning AI into practical tools for smarter forms, stronger customer experiences, and less busy work. We also explore the future of AI as an everyday co-pilot, whether you're growing your business or trying not to ruin 30 lbs of barbecue. Let's get into it.

Well, welcome everybody to another episode of The Good Neighbor. Today, we're joined by Thomas Kubeck from OpenAI. Welcome, Thomas.

Thank you. Thanks for having me.

I want to clear things up before we start. My mustache is newer, so I got it from him.

Yes, but yours came along the better.

No way. I just really like the way that it looks on him. I'm just being a little jealous and everything. I had to do it myself, too.

Sure.

But thank you for being here, man.

Yeah, thanks for having me. Really excited to chat today.

Oh, yeah. How was your day today? I know that you flew in.

Yeah, things have been great. I flew in to San Francisco on Monday. I've got a lot of customers here, so I manage our book of business and some of our digital native customers.

Uh-huh.

Many of those happen to be in San Francisco. So anytime I can make a trip down, get to talk to a lot of folks, that's always a lot of fun. It's most fun to engage with our customers and partners in person, and I always try to stop by Jform when I do.

That's awesome, man. I love having you here. This is our second meeting, I think, in person, which is great. We had a great time in our first, as well. Always productive, always to the chase, and everything. I'm leaving more room to have fun.

Yeah.

And today you're in luck because it's a wonderful day here. Adds up to our nice view as well. So tell me a little bit more about yourself. Who's Thomas? What have you done so far, and what's new?

What have I done so far in my life?

In your life.

Hey, life's a funny story. So I'm very local to Seattle. I'm from Seattle, Washington. That's where I grew up. And for a long time, if you drew a 10-mile circle, I lived in that circle my whole life.

For a while, with some exceptions, in a bubble, so to say.

Yeah, in a bubble with some exceptions. But, you know, I grew up as a native there. I went to University of Washington in Seattle. I started my career at Amazon, who's one of our great partners as well, and was there for seven years. During that time, I got the opportunity during COVID to move down to Nashville, where I met my fiancée. We moved back together. And then relatively recently, I joined OpenAI and have been super excited about the opportunity so far.

Okay, nice. Life has been treating you well so far.

Hey, I've got no complaints.

That's awesome. It's good that you have been a part of this big industry with giants early on, so that you know what you're talking about. You've probably been in close quarters with a lot of other huge names as well. It's great to have you when it comes to partnerships and how to deal with people, right? There's more and more AI every single day as we're moving forward. It's still good that we have that human interaction from time to time. That's wonderful. I want to segue into how our relationship started. I know this for a little bit, but just want to hear from you first.

Yeah, so I stepped into the Jform relationship relatively recently, but the OpenAI and Jform partnership, I think, has come a long way over a long period of time. I want to give Jform a lot of credit. I think they're one of the AI visionaries in their industry, and y'all are very focused on being early adopters. So I think where the relationship started, when I look through all of the notes over time, has been really around y'all driving innovation in your product, right? And really trying to make it easier for your customers to solve problems. It's, 'Hey, I need a form that does X, Y, or Z. I want analytics against this, and I don't want to be a specialist in form creation.' I am a small or medium-sized business. Maybe I'm a dentist, maybe I'm a church, maybe I'm some type of other organization.

Yeah.

Where that's my core competency, and through using your wonderful tools, which sometimes use our tools as well, you're able to allow them to get the information they need so that they can operate their organization more effectively. I think that was the start of the journey.

Mhm.

I think that it has blossomed with so many features that y'all have added and our continued support of those features. And then also, I think you can kind of make it into two buckets: how you go to market and how you support your customers. I think that's where it started. I think we're also seeing a lot of acceleration of Jform as a compilation of individuals, through Codex and ChatGPT work and some of the acceleration y'all have been able to experience there. I think those are the two frontiers that we're pushing together.

That's wonderful. I couldn't have put it better myself. As you said, we're more and more getting involved in making AI a part of our product. Nowadays, everybody somehow wants some sort of AI feature from the old version of doing things. After coming up with Jform AI, it was super easy to create forms with Jform before, too. Now it's a matter of seconds. It's just going crazy. The world is going crazy, but it's good. I'm not complaining. It was good to offer Jform inside your services as well, so people are having a great time chatting with OpenAI, getting their stuff done, automating their day, and they can also create a form right inside OpenAI if they want to.

That was awesome. And we are also using OpenAI services in our daily activities. Our team, and me personally, I cannot see myself not using Codex on a daily basis. I'm not blindly relying on OpenAI activities; it basically gives me the team that I need to see a GTM activity from top to bottom.

Mhm.

Creating content, everything. And I remember the first time we did a hack week at Jform during COVID times.

Mhm.

With Sam Altman, actually, back in the day. One of our VPs emailed back and forth with him. We did a hack week, and I remember the stuff that blew my mind back then now seems so mundane.

So, yeah. Now the stuff that we can do is like the Will Smith spaghetti video, right?

Yeah, yeah.

That's a wonderful thing. Every year they create one. The development is crazy. But generally...

That's a good segue. We're seeing development, right? We're seeing incredible speed; it's going crazy. How are you looking at it from where you sit? You're dealing with software companies on a daily basis and their use of AI. What's the trajectory looking like? I think it has its answer, but I want to hear your thoughts about it. It's pretty obvious.

I think that trajectory can look in a couple different ways. If we look at broad measures in terms of outcomes that people are starting to get, and the investment that people are starting to make, it's always up and to the right. As we're progressing, this is growing bigger than you or I, or all of San Francisco combined. I think what's really interesting is making sure that what you're doing is tied to outcomes. That's kind of the next phase. We went into token maxing, and I think that was a really good opportunity for companies and individuals to experiment, find hard problems to solve, and create creative solutions. Now we have to take this technology that makes anything possible and everything so exciting and make it boring, right? Draw the line toward making things useful, productive, and in the background as well. The big paradigm shift I anticipate is instead of working with AI by always delegating to it and making a lot of actions, can we bring you back to yourself and make it more about connection between people, with delegation and anticipation and a constant presence from AI that supports you? From an individual standpoint, I think that's kind of the future. Many of our customers at OpenAI are software providers, and they're providing some of the same experiences to their customers as well.

Okay, and then there's no quick way to see how it all ends, right? I feel like maybe not in our lifetime, but I wonder when the human element is going to get out of the picture completely. Do you think that's going to happen, or is there always going to be a human element somewhere out there?

Well, it's interesting, right? I work in a go-to-market organization, and most people consider my role sales. I would like to think I'm a person that's important for you to have a relationship with so that you feel like you can trust me.

Exactly. I wouldn't want to talk to those weird AI things. Definitely, I pick you every single time. But that's me. What about the next generation?

You know, I don't have a great, super firm point of view, but I think as a company we're more focused on empowering people than just replacing them. We want people to focus on extending their core competencies and having limitations taken away, not on just replacing them with nothing for humanity to do instead.

Right. What about the adoption you're seeing from these companies? Are they being hesitant? Security and privacy might be concerns, or competency, right? Just being stubborn and old? I'm counting myself as one of the guys, but I love technology and want to be active in it. How's adoption looking? What are some departments that are adopting AI faster, whereas some are more hesitant, and why? Is that within a company or your general views about working in OpenAI?

So at OpenAI, it was developer-first, right? Our research teams and developers were the ones using the most advanced tools all the time.

Early adopters.

They were the early adopters. What's been really interesting to see is that now agentic tooling is used by almost 100% of OpenAI. I think over 99% of the tokens generated within OpenAI for internal usage are via Codex or ChatGPT work. Sales was actually one of the fastest adopters. If we look at who adopted the more advanced tooling internally, we saw a ton of progress in sales. Sales is very easy to measure in terms of input and output, and is a very efficiency-driven task. There's always a mission that's easy to draw a line toward. For me, Codex allows me to spend more time with my customers.

Right.

There's no way I could do things like this and have these opportunities if I didn't have ChatGPT work supporting me in the background. Other leading facets we're seeing are on the financial side. Sara Fryer, our CFO, does an awesome job of talking about how we were able to rebuild an AI-native organization and how much better we are at sourcing information and closing our books more efficiently. That's been a huge transformation for us. Marketing is something we're really excited about, too. With the potential to create new assets quickly, ensure consistency at scale, and interact with customers in a more personal way, those are maybe the first three lines of business after the traditional technology or engineering groups.

Right. I always see these ratio things everywhere on my LinkedIn and stuff, as if it's similar to the fact that a huge portion of the oceans are still undiscovered. Is it the same when it comes to OpenAI adoption? Is there still a low amount of people actively using it, like yourself or myself, and a huge part of people yet to discover what OpenAI can do? I'm talking about regular people.

Sure.

I think it's really easy to get caught up in our bubble. OpenAI has broadly observed a distance between the amount of output tokens generated by frontier firms—firms seeing the outcomes of AI, perhaps loosely AI-pilled and really investing—and other firms. They are generating about 8.3 times more output tokens than other firms.

Okay.

So they are seeing significant usage. I think you're right, though, because the vast majority aren't frontier firms.

That's what it's like to live...

Right.

...on the frontiers. You're a little bit out on your own. Earlier you made a comment about enablement. If we're going to bring this AGI technology to everybody and evenly distribute some of the outcomes, we have to make sure everyone is able to participate. There are so many different levels to that, of course, but some of it is direct enablement around the tools. It's easy for me to look around the OpenAI office and say, 'Oh, everyone uses these tools. Everyone knows this. Everyone is chronically on Twitter seeing all of the updates.' But it's really important to look outside of that and say, 'In real applications, in real enterprises that are more traditional or maybe more legacy, we're giving them the opportunity to experience the technology, learn with it, and grow with it.'

What's the best way to reach out to those people? What do you think? Because they're not consuming media that much, right? I always feel like it's the same as our parents using smartphones or being on Facebook or whatever.

Mhm.

Yeah.

They were like—my parents were super hesitant about using smartphones and everything. And now they're doom-scrolling like there's no tomorrow. This is a real story, and I just don't know how it happened. I believe AI will become more common. I remember showing my dad the output of an AI; he was impressed by it. I don't know if he's still using it, but I think the medium is still yet to be decided. What's the best way to go out there with these people?

I think the number one thing that we can do is establish trust.

That's right.

I remember the first time I saw ChatGPT. I think it was 2022. I was sitting in a Taco Verde in Austin, Texas.

You remember it as if it were a first date, man. What was OpenAI wearing at that time?

I don't know. I do know that I met my fiancée on November 3rd in Nashville.

Good.

At the Whiskey Shot. So we've got the second one as OpenAI meeting.

Yeah, and so the second one is OpenAI. I remember being with my technology specialist at the time, and we were using it to generate SQL queries. It wasn't right. We could see how it was dynamically thinking and how this technology was going to change the world. I was like, 'I'm going to work for that company one day.' So I had to work my way there.

But...

Yeah.

I think that...

I didn't trust it then to deliver on critical workloads for me. Part of it is perception, because that's the first impression AI got to make on the world. Some people think we're still there.

Yeah.

There are some people who don't believe AI can do math because it struggled with basic arithmetic three years ago.

Right.

Now we're solving novel math issues. I think there's a lot of work we have to do on building trust with people that they can rely on the outputs. The other element that was an initial challenge is that the models were limited. Now I would say the models are fantastic and continuing to accelerate. The second element was context.

Mm.

Right? I always like to liken it to this: these models keep getting smarter. They started out as a middle school student, maybe got to high school, a college degree, a PhD, and so on. They're getting smarter and smarter.

Yeah.

But when you have that fresh MBA grad or fresh PhD show up to your office on day one, usually they're not very valuable because they don't know your business or your context, what you're working on.

Exactly, they learn.

As we've gotten better at supplying context with connectors, like maybe the Jform connector or any other kind of data repositories people are plugging into, that's where it's become more valuable. In your personal life, connecting to your email, calendar, or inbox of any kind—that's where it becomes more valuable. I think that's one thing we can enable the world to do more as we continue to work toward it.

You're absolutely right, because I always give this analogy when I'm talking about it. I'm a beginner guitar player, right? I learned it from YouTube. You pick up the guitar and start trying to do something with it. If it doesn't sound good, you toss the guitar away. People always think, 'Am I immediately really good at it?' If not, they're not interested. What the OpenAI experience did for me was, during our Codex Hack Week, a lot of people who were new to AI had the guitar in their hands.

Yeah.

Yeah.

They actually started playing a song from top to bottom, which made them want to go in and discover more. That's what it did for me, because I had a fully working product with a great interface in my hands in a matter of minutes.

Yeah.

And I was just talking to it.

Yeah.

So that's where the motivation comes from, right? The small wins.

Yeah, the feedback loop.

Yeah, exactly. You get those small wins quite a lot with AI help. That's why I like it. Now I can play the guitar quite well. I didn't have any help from AI with that, but it's the same, and I feel really good about it. I feel it's going to change the culture.

Yeah. Okay, well, I'm going to put you on the spot here. What's something in your personal life that you used AI to learn? Something you did for the first time, or a process you couldn't have worked your way through on your own?

You know what? I'm trying to learn about the legal aspects of being a homeowner in the US, for instance. Since I just relocated here, there's a lot to learn about you guys, man. I'm learning new stuff every single day, learning about the city of San Francisco in general and how to navigate myself in these situations. Before that, I had no one to trust, no one to get answers from.

Sure.

So I would say not maybe how to play chess, but how to survive in this life, basically. Recipes are always a personal way; they're just really helpful.

Yeah.

But what about you? What have you done?

The most fun example for me is related to recipes. I recently got a Traeger. I'm very excited about it. I've been doing barbecues for some of my friends and peers, and it's been a really helpful co-chef. It can guide me. I smoked 30 pounds of pork this weekend, and I wouldn't have known whether I could do it on my own. I can talk to it and say, 'Hey, have I bought enough meat? Do I have the right rubs? What am I not thinking about?' It's a relatively simple event to put together, but we've got busy work lives. I really had to get this right. I was hosting a lot of people.

I really got to get this right. You know, I was hosting a lot of people over.

And then getting the feedback loops, I think, is what we talked about. It was like, 'Hey, here's an image of the ribs sitting on the grill. Does this look right? Here's the temperature over time.' It's able to take all that context in and give me a quick feedback loop: 'Hey, you're on the right track.' Perfect, right? Although the AI isn't there physically, it really feels like it's sitting with me as my assistant to help me do the things in life I want to do. I want to share those meals and experiences with people. And I'm sure you want to be a homeowner and save...

Yeah, those...

Yeah, we'll figure it out. We're figuring things out, you know, just getting some good feedback, for sure. I want to go back to our partnership a little bit and how well we're doing together—Jform and OpenAI working well together, providing great solutions to mutual users. At the same time, you and I have been working on a lot of GTM activities, our relationship, and getting Jform to use more OpenAI as a company.

Mhm.

In your past experiences, including our experiences, what's something that makes you want to go back and work with somebody again? What makes you think, 'Okay, let's knock on this door once more because we've been doing this really well'?

Yeah, I think there are a few things. Obviously, it's more fun to work with people who lean in. When we're both excited about something, that's really helpful. It's also a lot easier to work with customers who have a really clear vision. Y'all have a clear vision, and that makes it easier because I know what your North Star is and I can help you get there. It's also helpful when we think about working with customers: who can lean in, move quickly, wants to be on the frontier, and can give us feedback? If we're not doing something well, we want to know how we can improve and do things better. Jform using our products and telling us where the rough edges might be is super valuable for us, and makes us really excited every time we get the chance to knock on Jform's door about new opportunities to work together.

That's great, man. That was wonderful to hear.

Me too. Just getting that feedback from anybody you work with rather than talking to a brick wall. Getting positive or negative feedback, whatever. When I said negative, okay, I'm going to segue. What's something you've seen as a failure example that people or listeners can learn from? Failure stories have always been better teachers than success stories.

Yeah, I think it's easier to take away learnings from them. If we look at the 2024 to 2025 timeline, it's funny how quickly things move. It was like, 'Just use AI.' That was the answer. I don't think 'just use AI' is the answer to every question. It's how to use AI to solve the question.

Yeah.

People looked at too high of a level. Really, we should look for problems to solve, understand what it looks like to solve that problem, and work backward from it. There are two elements to this. One, the technology has improved, so we can solve new types of challenges. But also, because the technology is improving, we have to look at problems differently. What a good solution looks like doesn't mean what it used to a year, two years, or three years ago. So I think it's about finding the right problem, thinking about how we solve it in the most modern way, and then executing. People get stuck when it's too high-level: 'Just use AI for everything.' Then it doesn't become valuable because it doesn't solve a problem.

Okay, yeah. Let's close out with the lightning round, shall we?

Sure.

Thomas, I'm really curious to get this answer from you: what's the weirdest AI use case you've seen?

When I look at the industry overall, I think there are some really interesting use cases on the social side. As a society, we haven't narrowed down how we want agents to drive interactions on behalf of people. Those are the ones that always make me stop and think about the impact. Everything we do every day is monumental, but so much of the value I see in AI is bringing people closer together. I'm always interested to see what's the right way to insert AI between people. Maybe using AI to find the right people to bring together is the right parallel. That's something I've always found really interesting. I'm not sure if we as a society have the right answer yet.

Yeah, not yet, right? Still in the prediction sense. Well, as I said, prediction from a year from now: what do you think? What's going to happen? How's Will Smith going to eat that spaghetti a year from now?

Well, it's funny. The Will Smith spaghetti video is getting harder and harder to tell apart year to year. One year it was looking like an alien eating like crazy, and now it's like, 'Okay, is this AI?' We're getting to that point. From a perception standpoint, what an individual model can do on a singular task, I think the gaps may become smaller for simple use cases, like generating a video.

Yeah.

I think we're really excited about the future of models and how we can challenge long-standing problems that require a lot of reasoning and more. But if I think about how I'll interact with AI, I used to use it just to answer my questions; it was like research. Now I use it as a co-pilot to help me triage and start all my tasks. In a year from now, I think it's going to let me really focus on the important things, like being there with my customers and spending more time with them, and doing less analysis and reporting. It'll just get done for me in real time, and I'll have something I can really trust for that. I think that's the future.

That is good. Let's listen to this podcast a year from now and see how right Thomas was. Or maybe we weren't even ambitious enough.

Maybe.

Okay, one final question, as I always do in a Good Neighbors podcast. If you think you have a person who would be a good fit as a guest on this podcast, who would that be? You want to pass the mic to?

I do want to pass the mic. I've been very fortunate to work with a lot of really special people over my career, and I wanted to bring someone I thought had a really strong point of view on AI overall. I wanted to nominate Michael Prince from Amazon Web Services. He's a senior generative AI specialist there. I think he'd be a great guest for you in the other chair. People have lots to learn from him.

Michael, I guess. Okay, there you go, Michael. You have been nominated. Hopefully we'll see him here as well. Thomas, that was amazing.

Yeah, of course.

Always great to have you in San Francisco.

So, if you have nothing else to add, that's it for our Good Neighbors podcast. What are your thoughts? Did you have a good time? Hopefully, say yes.

It is always...

It is always fun to have conversations like this, and I feel like it's something you have to set aside time for: reflection on all the changes. It's so easy to focus on Monday, Tuesday, and the next meeting I have. Pulling out and reflecting is a really good exercise. If I have anything to say, it's thank you for the partnership. I think this is such a unique opportunity for me personally and for OpenAI to participate in.

Wonderful to hear.

And I think we really want to continue investing in the partnership.

Thank you so much, Thomas, again. That was another episode of Good Neighbors. We had Thomas Kubeck here joining us live from our studios in our San Francisco office, and we'll catch you on the next episode.