OpenAI aims to disrupt app stores with new ChatGPT features

Key to this strategy is the ability for ChatGPT to suggest relevant apps within conversations, allowing users to connect and operate them directly within the chat interface. This is supported by an expanded plugin architecture that now includes extensions, enabling developers to build interactive panels. This approach aims to bring functionalities previously accessed through standalone web or mobile apps directly into ChatGPT.

Developers can create AI-native applications for ChatGPT, mirroring the development process for web or mobile app stores. This offers a new distribution channel as AI chat becomes a primary discovery mechanism. The "Sign in with ChatGPT" feature allows users to leverage their existing AI allowance across these integrated third-party applications. OpenAI has 16 launch partners for this initiative, including Cognition’s Devin and Notion, with plans to expand.

Demonstrations showcased how new apps, such as a meeting assistant, could function within ChatGPT, offering features like automated note-taking and summary generation. Design tools like Figma and Adobe were also highlighted, allowing users to work on projects directly within the platform. These applications can be shared through lightweight websites generated by ChatGPT, enabling collaboration with colleagues who can sign in with their own credentials.

OpenAI also detailed improvements to its plugin review process, aiming to provide developers with more transparency and control, akin to Apple's App Review. Furthermore, an enterprise app marketplace has been launched, featuring over 30 partners like Adobe, Salesforce, and Palo Alto Networks, allowing eligible customers to use their OpenAI commitment towards approved partner software.

The introduction of autonomous AI agents, known as Dots, further pushes this vision. These agents operate in their own cloud infrastructure with dedicated browsers and can utilize connected apps to perform tasks on behalf of users. This contrasts with the previous model where users had to manually download and manage various applications. Users will instruct an agent on a desired outcome, and the agent will handle the execution, potentially connecting to other AI models or applications as needed. These Dots can proactively offer assistance or handle forgotten tasks, always with user approval, and can connect to OpenAI's ecosystem of over 4,000 apps.