ChatGPT Apps SDK: What it means for your business

OpenAI announced what could lead to the biggest shake up in distribution since the App Store. It's own App Store?

OpenAI announced a series of new developments this week. Some of them have lit the internet on fire. But as with all of these kind of announcements, you have to take a step back and think about whether they’re actually significant or whether it’s more noise. Let’s break down one of the announcements that hasn’t received half as much attention as I think its worth: the ChatGPT Apps SDK.

What is the Apps SDK?

Essentially, the Apps SDK enables any business that has a mobile app to integrate that app natively within ChatGPT. Apps can be accessed by ChatGPT users by tagging it in conversation i.e. “Canva, turn this doc into a power point template”, or by having it suggested by ChatGPT based on context i.e. “would you like me to use Canva to turn this into a deck?”

Apps SDK = the new App Store?

This play is a carbon copy of the App Store model, originated by Apple, and the Skills model adopted by Amazon Alexa. We’ve seen this playbook before. Is there anything different this time?

Well, maybe.

You see, what the App Store did for Apple was thrust it into a position where iOS became THE distribution platform on mobile, in the same way as Google is (for now) THE distribution platform in the browser.

If you want your app available for your users, you have to go through Apple. And, as long as Apple has access to billions of users, what choice do you have?

The same was almost the case for Amazon Alexa.

What OpenAI MUST learn from Alexa’s failed skills store

Amazon had the fastest selling hardware ever. Faster than smartphones. It created a brand new product category and a brand new distribution channel. Brands hurried to deploy their services via Alexa from 2014 to 2022.

The problem? Discoverability.

It turns out that a headless interface with no screens made it incredibly difficult for users to understand what it can do, let alone what apps were available. Couple that with the intent-based NLU technology under the hood and it meant that, unless you asked explicitly for the specific skill you wanted, you’d never find it. And Amazon couldn’t help much because it didn’t understand what each of the skills in its skill store did in enough detail to be able to recommend many of them.

For app stores to be successful, you can’t just have distribution i.e. potential reach. You have to have discoverability.

After that, of course, you have to have engagement i.e. the apps have to work. They have to be good enough for people to return back to use again, and again, and again. Alexa’s problem was that there was so much inconsistency. Brands didn’t deploy useful stuff. They mostly deployed experiments and tests.

See, if the app isn’t useful, then no one has a reason to return.

And if you don’t have returning users, you can’t complete the final piece of the puzzle and introduce monetisation.

This 3 stage approach of discoverability, usability and monetisation is the process that unlocked the App Store for Apple and gave rise to the Uber’s and Airbnb’s of the world. So fundamental is this three-staged approach, that I had this exact conversation with Jo Jaquinta about Alexa in 2018; a conversation that essentially predicted the eventual decline of the skills store.

Why is this important? Because OpenAI cannot make the same mistakes with the Apps SDK as Amazon did with Alexa. It has to get it right.

So what’s different this time?

What sets the App SDK up for success?

Well, ChatGPT has significant adoption. 800m users in three years. The fastest adopted software ever. That’s more than 10x Alexa, which had 60m users after 3 years.

So OpenAI and ChatGPT are significantly further ahead in terms of genuinely competing to become the next distribution platform.

Secondly, the technology is much better. Having LLMs under the hood means that the level of understanding is far greater. Its ability to understand deeper context is second to none. And, having agents behind the scenes that are tracking the conversation, understanding what apps are available to solve which type of problems means that it’s far more likely that you’ll be recommended the right app to solve your problem at the right time (ima making an assumption that this is how it’ll work).

This means that OpenAI could crack the discoverability problem.

Third, the form factor quite frankly works better. Having a conversational UI with a GUI means that the experience of using the tool is more intentional. You can leverage widgets and maps and date pickers and all kinds of components that make the thing easier to use. This, coupled with the technology of generative AI, means that there’s a far greater chance of creating experiences that actually work. If you can do that, you can build repeatable use cases and repeat users.

Now, you have stickiness.

The monetisation will come eventually. OpenAI already has the Agentic Commerce Protocol which lays the foundations for commercialising app interactions in ChatGPT and there’s a proven blueprint for how that can be monetised a la the App Store.

The final thing the Apps SDK must get right: openness

The final lesson learned from Alexa was that closed platforms don’t move fast enough to satisfy the innovation requirements of the market. The amount of frustration coming from the Alexa community in the years leading up to the eventual decline in momentum was real indeed. And much of it stemmed from the fact that all features needed to be developed by Amazon first and released into the platform to be used. This meant that some things just weren’t possible to do.

OpenAI should heed this lesson and consider how it keeps a hold on standards (to make sure quality apps are integrated) whilst fostering innovation (allowing custom development of features).

So what does that mean for you and your business?

Well, genuinely, there could be some great opportunities to capitalise on OpenAI’s distribution here. A few things remain to be seen however, which as whether it’ll be a new customer acquisition play or an existing customer engagement play, whether your specific app, business and use case will be supported in the short term, how effortful or effortless it is to deploy these things, how you’ll manage and maintain quality, how you’ll gain visibility on the interactions so that you can improve them (also a challenge with Alexa skills) and many, many more questions remain.

What is certain, however, is that there’ll be a lot of experimentation and learning happening over the next few months.

Your AI Ultimatum is whether you apply to be involved now or hang tight and wait for it to play out.