[ivan_gammel]

>Other agent use cases: OpenAI doesnt have a surface to build these on. Google has chrome

ChatGPT app is their Chrome. Large consumer base using chat on daily basis can expand to prosumer and to enterprise. They build an emotional connection to their customers that has the vibe of iPhone.

[liamconnell]

Yes but its not relevant to the agent use cases (which are mostly about interacting with external systems). So agents built by Microsoft (Copilot) can natively interact with Office files in Sharepoint, and the Sharepoint product team can build to enable this in special ways. OpenAI has to use the APIs and deal with rate limits, speed issues and other limitations.

[ivan_gammel]

Do Oracle, SAP or Walmart invest in their own models or they build integrations? There are lots of companies which don’t have other options but to work with 3rd party LLM vendors. OpenAI, Anthropic and Mistral are better positioned for this than others.

[hobofan]

A large portion of those customers buy those models from their preexisting cloud vendors: AWS, Azure, GCP.

The customers are also price sensitive and are for many use-cases largely fine with last-generation model performance if that means they are cheaper. With that, there is little moat for the model creators, forcing a race to the bottom for model licensing, and the biggest chunk of the profit being captured by the cloud providers.

[throwway120385]

Oracle, SAP, and Walmart already have teams of people managing the data they have in-house. I find it hard to believe that those people can't pivot to working with an in-house LLM under the right leadership.

[ivan_gammel]

Data landlords have no reason to invest in their own models when they can license their content to all vendors and offer integrations with them. Their business models are enhanced by AI, not threatened by it. For Google and Meta it’s different: they offer entry points to data and chat apps are direct competitors. So chat apps with good integrations can be as big as largest ERPs, for example.