Introduction

My name is Zach Larick.

I lead the global SE team at OpenAI.

The Role of Solutions Engineers

We're called Solutions Engineers. So just very quickly, it's a team of folks that works with a wide variety of businesses of all shapes and sizes to help them understand how they can benefit from our technology.

Expanding Access to AI

uh but something that i'm like personally passionate about is helping the like 99 of the world that doesn't have access to chat gpt realize how they can accelerate their own workflow productivity and accelerate their dreams using this technology um 99 is probably a bold statement i would say 92 of the world has never used an ai utility which is which is wild right like we all how many of you are from the bay area

Most of us here, I would say within the 30 mile radius, you have 99% of the people that are working on AI. So we're definitely in a bubble here, right? And we have very close proximity to this technology.

Understanding the Current Usage

I think it's really important to understand that most people don't have access to it or they've never been exposed to it. And even those that use things like ChatGPT, and I honestly don't care if you use ChatGPT or Cloud or any AI utility, they're only interacting with it for like less than 10 minutes a day, which means they're not really using it or benefiting from it. which I think is surprising.

Opportunities and Learning

So there's a huge opportunity and there's two things you learn there. One is like there's a huge opportunity to extend this technology to a lot of people so they can benefit from it. And the second is if you learn this technology, you learn to use it well, it can be a clear differentiator for you.

As you saw in the last presentation, like you can have a very small team that can accelerate big outcomes or you can be an individual entrepreneur who can accelerate your dreams.

Personal Reflections

And Jeremy and I were talking about this earlier.

We both have children. I have a 10-year-old and a 7-year-old.

I live like 20 miles across the bridge.

And we think a lot about, what are you going to teach your children? What job are they going to have in 10 years? What jobs are even going to be around? And we think about that a lot at OpenAI.

And of course, I think about my children, even myself. And frankly, I think the one thing that I am certain of is that there will be more entrepreneurs in 10 years than ever before. And today, it is much easier than it ever was to start a business.

Demonstrating the Power of AI

And so today, I'm going to walk you through. It's going to be a 10-minute demo, very tight.

I appreciate if you clap or laugh at my jokes. There aren't many throughout.

All right, there we go. One, check. We got it.

Taking an Idea to Execution

thank you thank you all right that's a good start uh but we're gonna walk you through the process of taking an idea doing analysis on that idea uh prototyping an actual product that you could take to market and coming up with a plan for how you actually execute that and if you think about that in the context of 10 minutes if you can do that in 10 minutes think about how many business ideas you can spark and how lower the barrier of entry is for any average entrepreneur it doesn't matter where you are with access to a web browser to start a business and take it to somebody

raise money, hire a co-founder, etc. So I'll take you through that today.

Understanding the Audience

But before we get started, you can't see anything here. Can you raise your hand like how many people have used ChatGPT or any sort of like AI utility? Everybody here. Like how many would consider yourselves a power user? Excellent.

This is the toughest audience that I probably have ever. So I apologize if a lot of this is repetitive.

This is one of my favorite workflows. And the stuff that I'm going to follow, I think, is just fundamental to be part of your tool set. But please hit me up later if you want to go deeper on any topic.

Applying Technology to Traditional Industries

So a familiar screen and remember the story is we're starting with an idea here and this is gonna be kind of a weird idea. So I've been thinking a lot about like how do I start a business that applies technology to a traditional industry? And I've come across cold storage.

Does anybody know what cold storage is? You store things in the cold. Amazing.

But there is a lot of food scarcity in the world and it's a challenge of getting food to the places where it's produced to the places where they need it. So I thought about why can't we take technology and apply it to that problem and create a software product for that problem. So I'm gonna go through the sort of ideation phase for how I would come up with that idea.

Starting with Deep Research

The first step that I would take And this is something that lends itself to whatever your job is. If you're an entrepreneur, you work at a big company, anytime you're looking to learn more about something, and it's wild to think about this, but this is something that now, this is a few months old, but the concept of deep research now exists in Google and various other AI platforms.

It's only a few months old, but it's now like a fundamental component or part of your tool set. And I would recommend whether you're in sales, whatever profession you're in, you always start with deep research to do some analysis on the problem that you're going after.

In this case, I want to do market analysis for a software product for the cold storage industry. A very simple prompt. So I'll pop that in.

What the models are getting a lot better at across every frontier model is starting to guide you on the process. And I think a lot of people actually have a challenge today, and everybody's a power user in here, but more for the people that you're interacting with or teaching about this technology, is most people actually use ChatGPT for only one to two conversations a day, which means they're actually not using it at all. They're probably searching for something like they would on Google. The way that you get value from ChatGPT is to have a multi-turn conversation.

The models are getting better at sort of inducing that and enabling you to do so. And this is the first example of that where it's saying, okay, to help with market analysis for a software product, what kind of software do you want? That seems like important details. Are you targeting a specific geo? I'm going to keep it simple here and say inventory management in the USA. And we let that go.

Agentic Applications

This is also pretty cool, and I know you all know this, but this is sort of one of the first real agentic applications where you could set it on the task and then go accomplish that task on its own.

It brings up one important problem, though. What do you do while the agent's working? It's kind of awkward, right?

Yeah, you get a coffee, you go get distracted in Slack, you come back. It's something I'm still really working on as a middle manager. I'm really good at delegating and then going on to the next problem, but not to technology.

So in this case, we're starting this research, it's probably gonna take on average from seven to 10 minutes for this sort of problem.

Does anybody remember the Easy Bake Oven? Yeah, you do. Okay, great. I'm a little old. So like my references sometimes fall flat. But you were going to put that in the oven and I've already got a cake that's ready to take out of the oven here.

Market Analysis Results

So we've got the let's market analysis for software just to show you what that looks like in the interest of time. I'm starting at the bottom of the page. This is tough. I'm sure, for the folks in the back to make out.

And I'm sure you've gone through some deep research queries as well. But the way that you utilize this, I think, is really important. You've got all of your citations and resources.

And this is something that's rather new. I mean, this is literally like three-month-old technology in its accuracy and able to cite the right resources. If you want more detail, you can go into these resources.

For the purposes of this conversation, we're just looking for analysis on which market to go after and if it's viable. And just trust me, we're not going to scroll through all this. It's viable. There's an opportunity here.

As I scroll through this, I get things like customer segments and need of cold storage. I've got all my competitors here. I've got major players in landscape.

I mean, this is something that we all know would have taken six months ago, a team of five analysts months and they would come back and it wouldn't be as good at this. We show this to executives all the time and they're surprised. If there is one job at risk, it is the research analyst because they can accelerate these things. They can probably do many more reports like this, but it is very high quality and very usable immediately.

So in this case, I've got an assessment or a sense for the market.

From Research to Prototype

I know I want to go after this, but what I want to do is actually create a slide deck for some of the competitors that are servicing this because that's going to be part of the pitch that I put together. So I'm going to do another sort of agentic task here, and I'm going to start it from scratch, where I'm going to say, OK, great. Sometimes I like to talk to it like it's a human.

I just want to stay on their good side. I think it's important to say please and thank you and just create that dialogue, make sure it's in memory. You just never know.

Say, OK, great. Thank you. Thank you.

It's a good practice. It's also just good like wiring for the brain.

Yeah, yeah. Just be nice to everybody. Great.

Creating a Competitive Slide Deck

Thank you. I want to go after cold storage. Can you create a slide deck for me on the biggest competitors?

Typing in front of people is always easy. in the cold storage software market, a massive market. And then I'm gonna select Agent here.

Has anybody used Agent mode successfully? Yeah, it's pretty cool. Also, it's crazy that everybody's used it and it's only a few months old.

So what happens when you kick this off, of course, is that it spins up a virtual desktop and now it has access to all these tools. It can access Code Interpreter, it can create slides. Again, this is one of these things, Easy Bake Oven style, I'm not gonna go into, I'm not gonna make you wait for this, because it takes a little bit longer.

Let's go into, actual result of what that looks like sort of deep research like where it will cite the resources that it came up with uh but we've also got you know this decent looking slide deck that i use as a starting point where i say okay i've got you know an agenda of the slide deck um can you all see that let's see if we can actually zoom that out So it's like a decent slide deck.

It's got detail here. I mean, it's probably could use a little bit of polish, a little bit of a love and design eye. This is better than a slide deck I could create myself.

So I'll take it, but you know, it's got like really important, valuable information that I can use that I'm gonna form into my pitch. So we've got the slide deck, we've got the competitive analysis, we've got the market research.

Developing a Prototype

The next thing, which is really important, and that many companies, as they start, actually raise seed money to actually develop a prototype. So they wanna have something visual that they can actually take to an investor.

Or if you're internal at any organization, you want an idea that you can show to a decision maker and say, we're actually gonna invest in this idea. Some companies are now saying, if you have a feature request, you better vibe code that feature request first and bring it to me.

And then I want to see it before you actually make the traditional feature request. So it has applications across a wide variety of business climates.

So we're going to go and create a prototype for this. So I'm going to say, okay, great. Thank you again.

You are so good. Yes. Now, Can you create, and this is where actually it's kind of an interesting thing with GPT-5. I don't know if you all have played around with it.

You know, it has, it's always hard to get used to something new, including ourselves. It's a very steerable model, as they say. It means that it actually follows instructions better, which we thought would be the best thing. But the previous models would take liberties. And now with GPT-5, it is much better at instruction following.

So you have to give it better instructions. So sometimes I have to be very explicit and I say, okay, so now can you create a specific react prototype, which is for lack of a better definition of programming language prototype. And I'm going to use tailwind CSS, which is my favorite CSS library, which is effectively a way just to guide the style of this this application.

And it must work or I won't make any money. And this, this you've got children and you need to make, exactly. I don't know.

Like that was a used to prompt in the past and say, I would all get fired if you don't do this right. And it would provide more detailed responses. I don't think that's the case anymore, but, um, I always like to add a little, uh, emphasis or hyper hyperbole. All right. So I won't make any money.

We're going to do that again. Actually. Sorry. I want to live demo here. I'm not going to do that with agents. Don't need to stop. React prototype of my inventory management product for cold storage. Let's let it stop here for a second. All right, there we go.

Prototyping without Technical Skills

This is really powerful for me because I haven't written production code in like 20 years. Yes, I'm pretty old. It's been a long time since I've written code that was actually used for production purposes.

And I would be completely lost to have to do a prototype like this on my own. I would have to go hire a designer, hire a programmer, give them the idea. We're talking like a multi-month iteration or iterative process to make this happen.

And now what this is gonna do for me, and I use this in my own work life as well, not just creating prototypes, is this will actually spin up the entire prototype end to end with workable components. Another thing that I use this for is for creating my own dashboards.

Does anybody use Salesforce? Have you ever used Salesforce? It's a pain in the ass, and I used to work for Salesforce. I love them, I'm grateful for them. But it takes like an hour and a half, two hours to create a decent looking 2010 era dashboard.

And so now, like one shot, I can say like, here's a data set, here's some cool components, and I use my own dashboards in ChatGPT and just upload data to them. So it's a great tool for that. But in this case, you see it's writing code. We're not going to wait for it because it is very detailed.

I'll show you what this looks like, but you can imagine how powerful this is, right? You don't necessarily have to. You have to have technical aptitude, but you don't have to have experience building a prototype. This can be your first prototype. And this is what we call a one shot prompt. It's just a few sentences with limited detail, and it's going to come up with a really good looking application. And I'll show you an example of what that looks like. Once it's finished, you can click preview. But let's go into already finished out of the oven prototype here. Click preview. And it looks pretty good, right?

I mean, you can make some various iterations here. I can actually go and interact with these components. I've got dropdowns.

I mean, it looks like, for all intents and purposes, like a prototype version of the application that you'd want to go raise money on with your business plan that you'd go recruit co-founders. Or even just for your family to say, like, I have this real idea. This is what it's going to look like. And you get people visually engaged with the concept in addition to your business plan.

So very cool. And also a massive lowering of a barrier of entry to starting a business, which I think is pretty cool. In addition to that, what you can do is iterate. So I could go and say, like, I want this to be green or brown. I want this to behave in a different way. It's all interactive and I can also share it with folks.

So if I have somebody I'm already working on this project with and if you can see this in the back of the room, I can create a link and we can iterate it on it together. So I think super powerful and especially for folks who are not as technical or looking to get into vibe coding, this is like your first step here. You really don't need technical aptitude to do it. You just need to know, have a design eye and have a feeling for what you want the application to look like, which I think is pretty cool. I would also say that I have friends who are designers and programmers who do this as well.

And it's not replacing those jobs. It's basically giving them a mechanism to create many different prototypes and iterations of that in a much more rapid manner. So they can iterate on a daily basis versus weeks or months, get feedback. And then they've got this working prototype that they can give to the engineer and the engineer can take it to production. So pretty cool.

But for this individual entrepreneur, we've got our prototype. We've got the competitive slide deck.

Building a Compelling Business Plan

We've got the semblance of the research. Now I want to pull this together into an actual business plan.

So I will pop out of here. And I'm going to start a new chat. I'm going to use some of the assets we already have. And I'm going to say I'm going to drag them in here. See if I can do this gracefully.

I only copied one of them. Bear with me. Here we go.

So we've got the slide deck, and we've got that market analysis. So with deep research, you can also save it to a PDF. Makes it easier to access or share. So I'm going to add that to the chat.

Pitching and Partnerships

Competitive analysis, market analysis. We've got the accompanying prototype. And I'm going to say, OK, I'm ready to pitch, my friend. Hope you are well. Can you? Can you please come up with a clear and concise investor pitch?

I also want to hire co founder and help me with company names. So again, this is something that would have taken me previously weeks. I would be in sort of analysis paralysis. What is the right pitch? Like what's the right way to put this together for folks?

So it's gonna go and take those two outputs that we created in the last 10 minutes. It's being very nice to me as well. So it's a mutual relationship we have. And we've got the concise industry pitch.

It also references the file that I uploaded. So you have the resources that you can pop into. We've got the solution outline.

So when you're going to talk to an investor, you can be very concise. Going to talk to somebody you're about to hire, you're very precise about the problem that you're trying to solve. You've got the pitch for the co-founder, right, real directly.

Like here's, we're solving one of the toughest problems in food and pharma, making the cold chain smarter, greener, and safer. Company needs ideas or name ideas. Who votes for cryo flow? No. We've got chill logic. Chill logic? All right, we're chill logic. It could be glatient as well, which is also a great idea. We can iterate on that.

Refining the Pitch Deck

And then it says, okay, next step, I can refine the pitch deck. And this is actually something that's pretty new, where it's going to continue to guide you through the problem that you're looking to solve. So it sort of infers the business outcome, which is like we've got the tight pitch deck, and then says it's probably going to want to create a slide deck for this.

So I say, all right, let's do it. and we'll create that slide deck. So a lot less generating.

I know that you all are power users. This is probably a real simple example, something that you've walked through previously. But I think what's important is to start to think about, as Jeremy mentioned, how you're going to put your ideas into action.

The Era of Low Barrier Entrepreneurship

The barrier of entry is lower than ever. I mean, another way to say that is there are no more excuses.

Like if you have an idea that you want to start a business, like there's no excuse in market analysis to determine if it's a great idea. There are no excuses to actually to building that applications to a state where you can actually get, you know, folks to help you take it to production. And there are no excuses to help you iterate on an action plan as well.

Real World Applications

And I've done this with my 10 year old son. We started a website for, it's called InjuryComfort.com, shout out, where we sell injury products.

And what we do is we actually sit together and we'll take a notebook out and we'll write down our ideas, which I think is very important. Like you don't want to eliminate your brain from the entire process.

But we write down those ideas and then we take a picture, we upload to ChatGPT and say, ChatGPT, give me a plan of action. Like what are the next steps? And we work together on executing that plan.

And we executed it faster where there's actually no work to be done. And he keeps asking me, like, what can we do on the website? And I'm like, there's nothing to do, man, sorry. So we got to go on to the next business idea.

Future of Entrepreneurship

So I think there's a world in the future where you have entrepreneurs where previously they can only start one business at a time, maybe one business a year, because that's all of the operational and technical overhead they had. Well, now you have entrepreneurs with a team of two or three that have many to hundreds of businesses that they can iterate on because all the technical infrastructure is abstracted away.

Conclusion

So at any rate, I know I'm over time.

Hopefully that was helpful.

Thank you all for your time and hope you have a good evening.