Emerging AI business models

This post began as part of a wider post about AI features but I’ve decided to split it out. I expect I’ll update this list over time but for now, these are the main AI business models that I experience most.

I hope this will be general and relevant enough to anyone - whether you’re an individual customer, a business accessing or providing work approved tools or an company (like us - Battenhall) that builds products on top of these models.

Come at me with questions, corrections, addendums and anything else that takes your fancy this Friday!

API providers

This is the playground of Google, OpenAI, Anthropic et al.

1. APIs

The classic growth enabler for an emerging market. Model makers sell access to their tech via APIs that software builders integrate into other apps. This is what powers AI features in seemingly every single app we use today (whether you asked for it or not!), because not many companies have the resources to build this from first principles. AI enables apps to do other things and that’s now a cost of doing business for those builders.

This is a good model for the customer (app builders) and end users, but with competition growing many models could do the same tasks and so there’s a risk of becoming a commodity. Plus, businesses throughout history love predictable recurring revenue rather than a fickle PAYG customer based.

An example from my personal life is the Runna app, where I pay a subscription for their running apps but they use AI as a small component to create post-activity analysis.

This applies to most B2B and B2C apps, with the notable exception of probably Google Workspace and to a lesser extent Microsoft Office365 because they either have their own models (Goog) or investments and special relationships into providers (MS).

2. First party apps

The phenomenal rise of ChatGPT proved the interest in a subscription model for AI providers, where they themselves build an app around their models. It creates a symbiotic relationship where they get premium access to the AI with a great user interface and the providers get valuable usage data to steer model development (increasingly important as the open web shrivels and training data becomes a differentiator).

ChatGPT and Claude are probably the prime examples here, with Google offering it too but also bundling Gemini into Workspace and their device offerings so they’re more diversified.

3. Wrappers, aka model margin arbitrageurs

There’s an overlap here with the API model, but in this early and fast moving stage there’s a lot of this about. There are business selling AI access directly but wrapping that in a service that allows them to skim a percentage in exchange for greater access.

In my software development world, I pay a subscription to Cursor and they give me access to a wide range of models. Their business model is based on giving me that choice but betting that I won’t use my full allowance and they pocket the difference.

This is a hybrid of the API and First-party business models but it’s only possible because of a) the access to models via API but also b) addressable markets that aren’t serviced by first party apps. These things are not guaranteed to last, however the rise of open models means that it’s not likely that the API market will close up due to monopolistic tendencies for a while

SaaS providers / B2B apps

Please don’t feel left out - I’m going to come back to this another time. For now these apps are largely going to be enabled by the API business model above. We’re not at the stage where the average company can do something bespoke and build their own models but that is likely to change as the economics, technology and resources evolve over time.

What have I missed?

I would love to hear your thoughts in response to this list so far. What should we include in future versions?