A "Once In 15 Years" Product Opportunity
How to Build a No-Code GPT Business in Under 10 Minutes
Open AI's developer conference this week may have gone under your radar.
It shouldn’t have.
Why?
They made one huge announcement:
The launch of a GPT App Store at the end of November.
That means developers can build & publish GPT apps to a captive audience of ChatGPT users - solving perhaps the biggest challenge for early-stage businesses:
Distribution (i.e. potential customers discovering your product).
To illustrate the potential of this app store, imagine what it was like launching on the Apple App Store in the first few months?
Rapid growth of audience (potential users), still few apps published (less competition) and, importantly, revenue to capture a part of.
Creating a GPT Assistant
These apps are called Assistant GPTs & the building them is very simple:
Define Instructions: What do you want the Assistant to do? In the example above, for example, I’ve asked it to advise aspiring entrepreneurs on how to come up with ideas for an Assistant to build
Model/tools: You can define which aspects of ChatGPT you want the Assistant to use. For example, the latest model? Maybe a bit slower & more expensive, but better results. Or maybe you want to integrate ChatGPT’s Code Interpreter for a business/tech-related Asssistant
Files: You can upload datasets to train your Assistant, such as your personal finance data if you were to build a Financial Advisor Assistant
Test: You can test out your Assistant by interacting with it through a chat interface
Publish (Available Dec 2023): Once you’re happy with it, you’ll be able to publish it on the public App Store with a single click, or keep it for private use as well
The REAL Challenges of Building a GPT Assistant
Building these Assistants is unbelievably easy.
However, that very fact creates a dichotomy:
They are so easy to build that everyone will build them — and build a lot of them.
For example, my team are aiming to build 10 Assistants to go live as soon as the App Store goes live (supposedly end of November).
That means building an Assistant that people actually use is going to be hard.
Very hard.
Here are some of the Assistant GPTs I saw featured in the announcement so you can get an idea of what’s possible:
But, there will be intense competition within each category; imagine how many “Personal Trainer" Assistants will go live within the first day or two?
That means you will need to be rigorous in defining solutions that have a good chance of 1) Creating value for an audience and 2) An ability (or path towards) capturing some of that value in the form of revenue.
To do that, it’s essential to understanding the context of these Assistant GPTs before jumping into coming up with ideas for individual Assistants. Why? If you’re not clear that people will discover your Assistant through search or category filters (e.g. Productivity or Education), then your Assistant may never even be discovered.
Some key assumptions I’ve researched so you don’t have to:
How to Come Up With High-Potential Assistant GPT Ideas
Now, to the hard part.
In a world where every forward-thinking business & aspiring entrepreneur is going to be launching Assistant GPTs left, right & centre, how to build one people actually use?!
Well, first, it’s worth starting that we just don’t know what is going to be valuable until we publish it & see how/if users engage with the product.
However, we can give ourselves a better chance of success by evaluating which ideas do - and do not - make sense.
To do that, I’ve put together a worksheet to qualify the 30-40 ideas my team have come up with:
How to use this worksheet? Simply fill in each column. Here’s how:
Problem: As with any product, you need to be clear what problem you are solving. Be specific. Be concise. For example, “manage your health is difficult (the problem) without multiple apps or a personal trainer is difficult (the cause)”
Niche: Every problem has a person suffering from it. Again, be specific here. In such a competitive market, your niche should be as specific as possible. Try also to think of somebody who the problem is really acute for. For example, your product is far more likely to succeed by being targeted at “first-time Product Managers in an early-stage startup” than for “all Product Managers”. You’ll be able to deliver more specific value & your offer will resonate more strongly with a specific audience. Remember: Facebook didn’t start by targeting everyone. They targeted a single year group at Harvard.
Stage 1 UVP: There will be two types of Assistant on the App Store. Those built using the ChatGPT interface (what I call Stage 1) & those that are custom, external apps using the ChatGPT API (Stage 2). An example of Stage 1 would be a simple “travel advisor app” that can answer your questions about, for example, where to go on holiday
Your unique value proposition - the thing that will help you stand out in the eyes of your target audience - will likely change or evolve as you move to Stage 2.
Why? Simple. You will have more capability to customise the user experience. However, defining a clear UVP at Stage 1 is crucial because this will unlock access to the market (i.e. people finding you on the App Store) & delivering enough value to them so they actually use - & hopefully come back to use - your productStage 2 UVP: This is where you’re likely able to turn a captive audience into a real business. Rather than just answering questions, your Stage 2 might be to become a “travel concierge service” that suggests & books flights, accommodation & things to do for you on an external platform - as well as store all those bookings for when you actually go on the trip.
Acute: Rate from 1-5 how acute the problem is. Is it a vitamin/nice-to-have (1) or a painkiller/necessity (5)? The more acute the problem for this niche, the more motivated they will be to discover, test out & continue to use your product.
Addressability: Usually it is a huge challenge to consistently get access to your audience (unless you have a huge advertising budget). With the App Store, as long as you define your category clearly & have a title that is easy to discover through keywords(e.g. “Personal Trainer GPT”), your target audience will find you
Defensibility: Rate from 1-5 how easy to copy your product will be. If it’s just a generic “travel advisor app”, it’s pretty easy to copy. In fact, they’ll be 100+ of them within the first few days of this App Store going live. If, instead, you’re able to upload unique, rich data your company has on thousands of travel reviews or previous trips people have taken, your suggestions will be a lot better (making the app more valuable & hard to copy). To be defensible, try to think of unique data sets you could use, or how to really tailor your product to your specific niche.
Revenue: Rate from 1-5 how much revenue potential your product has. Your costs are going to be low with this kind of business, as users will use your Assistant through their ChatGPT credits or subscription plan. The hard part will be working out how to charge your users in order to turn a profit. That might mean waiting until Stage 2 to deliver a more valuable product to charge the user for. It might mean looking at how much this audience tends to spend on similar products (for example, most people use travel products for free, so you might put revenue potential at 1/5. If it’s a B2B product, it’s pretty normal to pay a subscription of $10-20/user, so revenue potential could be 4/5).
Finally, you’ll get a score to get a sense of the potential of each idea.
My advice then?
Get 20-30 ideas down.
Select 5-10 you see potential in.
The difference is the App Store was purely a platform, whereas ChatGPT apps are more like future product features.
The more pertinent analogy is Amazon Marketplace and Amazon Basics: Where Amazon lets vendors experiment with ideas at their own risk, and then spots opportunities to add to their assortment of Basics products and undercut existing suppliers.
OpenAI did this recently with its “chat with PDF” feature, for example.
For me the OpenAI marketplace is nice for making a quick buck, or building a cool portfolio project - which is not a bad outcome, just not quite the chance to build a business...