Future Revenues of AI

Let's get something straight: every company's sole purpose is generating profit. Speaking of which, when the revenue declines, costs are rising or even both, these companies need to seek for new revenue sources. Often, they surge to find them outside of their current addressable markets, sometimes it is in their addressable markets. But what if the revenue existed all along before, it just streamed into a different pocket?

Background:

Lately, an interview with OpenAI's CFO Sarah Friar revealed possible future ambitions to take profit from output. Means if a company uses OpenAI's services to generate revenue or cost reduction, OpenAI could register ownership of parts of this.

She compares OpenAI's business model with a Rubik's cube. With rising count of pieces, the count of possibilities grows as well. Means as a company uncovers more and more ways to generate revenue, the more complex the company gets.

I like this analogy because it reflects the reality. The bigger the company, the more people working for it and the more products it serves to its customers, the more possibilities there are to uncover possible revenue streams.

Let's imagine the following scenario:

A company uses AI to find and scrape potential customer contact data on the internet. The AI even makes the first step and contacts the potential customer based on an priority algorithm created by the company. This process enables the data application team, to enrich their own databases with valuable free available data and the sales team to generate new customers to generate new revenue streams. Now OpenAI comes into play: they want a piece of the cake. They specifically ask to get a share of the revenue that has been generated due to their products help.

This poses 2 significant questions:

1. How to determine the part of revenue that OpenAI is entitled to ask for due to it's part of the process?
2. How does OpenAI's entitlement in the value-chain influence future revenue streams?

Legally, this is a extremely complex topic. Dividing the revenue that was possibly generated by past process optimization is a problem, that can not be easily answered. How much growth did OpenAI enable by providing the sophisticated model?

Revenue participation as payment for a service itself is not factitious, in fact, it is often used in different business models. Almost every payment processors business model is built this way. This approach overall is not new by definition but new because the demand is not transparent in it's scope. The payment processor e.g. takes 1,4% of all payments that has been processed - this is easy to predict and to forecast. But the specific approach OpenAI imposes is hard to calculable. What would be the basis of this calculation? The saved time in comparison to the old process? But can OpenAI know how long the old process took? How far into the future do the claims extend?

Logically, these decisions always end to in a favorable way for the AI providers.

Conclusion:

For AI providers, this is a sensitive topic as well. If they register partial ownership of their customers revenue stream, customers could decide that this implies a risk to future growth and use another provider or set up own LLMs.

Just imagine you buy shovels because you want to dig for gold and when you finally find some, the seller wants it's part of your profit.

Let's spin this wheel further, what could happen in the future?

Vertical steps of supply-chains/value-chains could merge together under the hands of few companies that control critical parts of processes. If AI is integrated everywhere, the providers of the AI are integrated as well.

It is mandatory to secure one's business model by building processes without ‘too far-reaching’ external dependencies. It is a major potential threat to a company's business model if it relies on an legally unsafe state.

Recommended Actions:

As a company:

→ Learn to automate processes without the extensive use of AI from external providers

→ Rely on your own resources like knowledge to handle process ownership

→ Build stable processes without using AI for complex problems

As a private person:

→ No need to change something if you only use AI personally

→ Think about moving to self-hosted solutions if you have the resources

→ If you plan to found a company in near time, think about process automation without excessive use of AI

Striking every problem with AI seems easy and fast to do, but in the long term, it does not necessarily strengthen your process.