GPT Store: The "Apple Store" for GPTs

Last week OpenAI announced the launch of their GPT Store. Let's have a look at what this GPT Store exactly is and how it can impact your business.

GPT Store: What is it?

First of all: what is the GPT Store exactly? For starters, the GPT Store is a marketplace created by OpenAI for AI apps and bots built with ChatGPT. It is a platform for developers and creators to publish GPT-powered tools, also called GPTs, for the world to use. The store tries to offer a variety of customized GPTs to all OpenAI users and currently already features a diverse range of GPTs developed by both OpenAI's partners and the community. More and more GPTs are expected to be developed over time as the GPT Store will be up and running. Moreover, OpenAI envisions that the store will evolve into a marketplace with 1000s of custom chatbots that are trained/prompted to solve highly specific issues. A leaderboard will then be used to rank the performance of the different GPTs on these different use cases. This leaderboard will rank all the custom bots and multiple categories such as DALL-E (for image-based problems), writing, programming, education, lifestyle, and many more.



So, how do I create a GPT?

It's important to know that the presence of multiple high quality and highly used GPTs will make or break the GPT Store concept for OpenAI. So they went all out in making the process of creating those GPTs as easy as possible for you. Let's look at the different steps into creating your own GPT.

First of all, you need to verify a domain for your GPT. For example, on the event where OpenAI launched their GPT Store, Sam Altman created a GPT with the domain: Startup Mentor. There are some rules to choosing your domain name but these are all quite straightforward. Some of them include that you can only use an organization's trademark if you are authorized to do so and that referencing public figures, profanity and harmful topics are prohibited. In other words, creating a GPT and calling it "GPT-5" might sound like a fun idea, but your GPT will be taken down rather quickly - someone actually already tried this and the GPT was taken down in a heartbeat.

Alright, your domain name has been chosen and now the real work begins: it's time to create your own GPT. This can be done in a very low code/no code environment. You can essentially just interact with GPT-4 and create your GPT through clever prompting. Barreling on the Startup Mentor GPT example, Altman simply instructed GPT-4 that he wanted to "help startup founders think through their business ideas and get advice. After the founder has gotten some advice, grill them on why they are not growing faster". After hitting enter, GPT-4 just starts thinking and creating the GPT for you, no coding required on your behalf.

Now, you're probably thinking: if creating a GPT is that simple, why would I ever use GPTs created by others and not just create my own GPT on the fly? Well, there's a catch. While GPT-4 is building your GPT, you can also alter some of the configurations yourself. Say, you have some lectures/ documentation on how to begin a startup; one thing you could do is upload those to your GPT to make sure it has access to that information. This can be done through an upload button. For those who are more familiar with Generative AI, this setup will essentially create a RAG system for you. In essence, the GPT will use the information that you have provided to give answers to the questions of its users. Apart from uploading files, you can also configure certain capabilities such as web search and image generation as well as add further instructions to the GPT through an instruction prompt.

Now, your GPT is ready for use! Interestingly, OpenAI is also investigating how they can incentivize developers to create these GPTs by giving some monetary incentives. While their monetization program is currently only being tested out in the US, GPT builders will be paid an amount of money based on the engagement of their bot. It is yet to be determined how effective this strategy will be and whether OpenAI will later roll this out into a wider public.

What's in it for your business?

Let one thing be clear: the GPT Store looks great and definitely has potential. However, the question quickly rises: "Does it also have potential for my business?" While we believe that most businesses will not need to create their own GPT in the Store, let's have a look at when you would and would not consider doing so.

First of all, why create a GPT when you can just browse and use the GPTs that have been created by others? Some of the GPTs that will come out in the future will undoubtedly be amazing, so make sure not to miss them. We say it's time to sit back and relax, scroll the GPT store once in a while instead of your LinkedIn and X feeds and make sure to try out some GPTs!

However, let's say that you are excited to build your own one and you have data that could be interesting for parties outside your organization to use and you are keen on generating some money from that. One example of such data could be data about past court trials and their outcomes. One could create a GPT that has access to all information of these trials and then host a GPT that has access to that information. If this GPT gets used by a lot of lawyers, one could expect to start making some money from it - however, note that the monetization is currently only activated in the United States. You could also experiment with making your GPT private and only granting access to it after a payment has been made.

Next, if you are planning to only use your GPT internally it makes little to no sense to develop your GPT in OpenAI's store. While the experience to creating a GPT might look very simple, it might turn out to be everything but convenient keep it up to date. Moreover, let's say you want to create a RAG system to chat with your own data. In that case you are probably already using either Microsoft Azure, Google Cloud or AWS as a cloud platform, so why bother creating your GPT on even another platform than the one where your data is already located? This would also mean that your potentially sensitive data will be stored at OpenAI in order for it to be accessible to the GPT. On the other hand, uploading a single file for the GPT to use is quite simple so for one shot analyses, it could be handy.

However, the GPT-store has some limitations e.g. what if you have a multiple files you want to update periodically? What if you want to customize your UI? What if you want restrict accesses? What if you want to add first-party data to improve the customer's experience? These complex questions will require a more custom approach, which is where a cloud platform can prove valuable. If you are interested in building your own internal GPT-store on the Azure platform, don't hesitate to reach out!