When building AI-driven workflows, the primary interface becomes the API, not the GUI. A tool's value is determined by its programmatic control. Consequently, a clunky UI with a strong API like Salesforce can be superior for AI integration than a tool with a slick UI but a weak API.

The next generation of software may lack traditional user interfaces. Instead, they will be 'API-first' or 'agent-first,' integrating directly into existing workflows like Slack or email. Software will increasingly 'visit the user' rather than requiring the user to visit a dashboard.

AI agents often default to "build it yourself" because SaaS products aren't designed for them. To stay relevant, SaaS companies must create agent-friendly CLIs, APIs, and even add hints in help text to guide agents through complex workflows.

For companies building AI agents, the key indicator of a successful customer engagement is the availability of well-documented APIs. These APIs are essential for the agent to take action and look up data, which directly enables a superior, elevated experience from day one.

Karpathy's home automation agent ("Dobby") replaced six different apps by interacting directly with smart device APIs. This suggests a future where users interact with a single agent, and software products must expose agent-friendly APIs to survive, as their bespoke UIs become irrelevant.

As companies integrate AI agents into their workflows, unrestricted API access to their own data is non-negotiable. SaaS providers that paywall or limit API access will be abandoned for more open platforms that don't hold customer data "ransom."

In a world where AI agents perform tasks, the value of a SaaS product is no longer its user-friendly interface but the robustness of its APIs. The core differentiator becomes the proprietary business logic, security, and data governance embedded within the API layer.

When a user wants their AI agent to have deep access to a SaaS tool like Slack and is denied, they can now use the agent to migrate to an open-source alternative like Mattermost. This creates immense pressure on incumbent SaaS companies to provide robust, open APIs or risk losing customers.

Standard APIs for human developers are often too verbose for AI agents. Notion created agent-centric APIs, like a special markdown dialect and a SQLite interface, by treating the AI as a new type of user. This involved empirical testing to understand what formats agents are naturally good at using.