Did ChatGPT Just Kill Zapier?

OpenAI's latest releases, including new app integrations within ChatGPT and the AgentKit for building AI agents, significantly enhance AI capabilities and disrupt existing automation platforms.

• ChatGPT now integrates directly with third-party apps, allowing in-app task execution.

• AgentKit provides a visual builder for complex, multi-agent AI workflows.

• These advancements position OpenAI to disrupt existing automation platforms and foster a new ecosystem of AI-driven tools.

OpenAI's Dev Day introduced two major advancements: the ability to directly use third-party apps within ChatGPT and AgentKit, a new visual builder for multi-agent AI workflows. These features streamline user interaction with various services and enable the creation of complex automated tasks, potentially disrupting companies like Zapier and Make. OpenAI is fostering an ecosystem where developers can integrate their apps and build sophisticated AI agents, with future monetization and natural language agent creation anticipated.

ChatGPT App Integrations

• 00:01:22 ChatGPT now allows direct interaction with popular apps like Spotify, Canva, Figma, and Booking.com without leaving the chat interface. Users can prompt ChatGPT to perform tasks within these apps, such as creating a Spotify playlist or generating a Canva flyer, and then make real-time adjustments. This feature, which uses an SDK for developer integration, is a more intuitive and integrated version of previous plugin attempts, significantly lowering the barrier to use.

• 00:03:02 The system streamlines workflows by automatically suggesting and connecting to relevant apps. For example, if a user requests a flyer, ChatGPT might suggest using Canva and then facilitate its creation and modification. This direct integration transforms ChatGPT into a versatile assistant, enabling a wider array of tasks and fostering a growing ecosystem of integrated applications, with future monetization for developers expected.

Introducing AgentKit

• 00:08:50 AgentKit, available in OpenAI's developer playground, is a visual canvas for creating multi-agent AI workflows, similar to tools like Zapier or Make.com. It allows users to build agents that connect to various apps and websites, leveraging Model Context Protocols (MCPs) to enable integration with any product that develops an MCP server. While currently limited to OpenAI models, this platform has the potential for extensive automation.

• 00:10:38 AgentKit provides templates for common tasks like data enrichment, customer service bots, and structured data Q&A, and supports custom agent creation using a node-based workflow. It features guardrails to prevent sharing personally identifiable information or detecting hallucinations, and allows for conditional logic with 'if-else' statements to branch agent responses. This early-stage tool hints at a future of complex, interconnected AI automations.

Disrupting Existing Platforms

• 00:08:50 The introduction of AgentKit poses a significant disruption to existing automation platforms such as N8n, Mind Studio, Make.com, and Zapier. While these established platforms currently offer broader model compatibility and extensive integrations, AgentKit's direct integration with OpenAI's powerful models and its potential for an MCP-driven ecosystem could eventually rival their capabilities. The ease of use and native integration within the OpenAI environment presents a strong competitive advantage.

• 00:14:44 AgentKit's development suggests a future where users can build complex automations directly within the OpenAI ecosystem, potentially using natural language prompts to construct entire workflows. Although not yet as robust as some established platforms in terms of external model integration, its rapid evolution and direct connection to core OpenAI technologies position it as a formidable competitor, challenging other companies to differentiate their offerings.

Future Outlook

• 00:14:44 The capabilities introduced by ChatGPT's app integrations and AgentKit point towards an infinitely scalable future for AI-driven automation. As more companies develop MCP servers and integrate with OpenAI, the potential for building intricate, multi-tool workflows will expand dramatically, allowing for tasks such as prompting an action, sending it through multiple tools, notifying team members, and sending emails.

• 00:17:10 A key prediction for the future is the ability to build these complex agent workflows using natural language. Users could simply describe a desired automation, and the AI would construct the underlying node-based workflow. This advancement would democratize agent creation, making powerful automation accessible to a much broader audience and further cementing OpenAI's role as a foundational AI platform.