Notion's most significant AI breakthrough stemmed from a surprising source: simplification. When initially exploring large language models (LLMs) and agentic AI, engineers at Notion AI experimented with advanced code generation, complex schemas, and extensive instruction sets. However, through trial and error, the team discovered that eliminating complicated data modeling dramatically improved model performance.

Ryan Nystrom, Notion AI's engineering lead, and his team pivoted to simple prompts, human-readable representations, minimal abstraction, and familiar markdown formats. This re-wired approach led to the release of V3 of Notion's productivity software in September, featuring customizable AI agents. These agents have quickly become Notion's most successful AI tool to date.

Nystrom described the improvement as a "step function," based on usage patterns compared to previous versions. "It's that feeling of when the product is being pulled out of you rather than you trying to push," Nystrom explained on a VB Beyond the Pilot podcast. "We knew from that moment, really early on, that we had something. Now it's, 'How could I ever use Notion without this feature?'"

The initial approach to integrating AI involved complex technical strategies. However, the team found that LLMs responded better to simpler, more direct inputs. This shift reflects a broader trend in AI development, where focusing on clarity and usability can yield better results than sheer computational power. The implications of this finding extend beyond Notion, suggesting that the key to unlocking the full potential of AI may lie in making it more accessible and intuitive.

Customizable AI agents represent a significant advancement in productivity software. These agents can be tailored to specific tasks and workflows, automating processes and freeing up users to focus on more creative and strategic work. The success of Notion's AI agents indicates a growing demand for AI tools that are not only powerful but also easy to use and integrate into existing workflows.

As a traditional software engineer, Nystrom's experience highlights the evolving role of engineers in the age of AI. The focus is shifting from writing complex code to designing effective prompts and user interfaces that allow users to interact with AI in a natural and intuitive way. This requires a deep understanding of both AI technology and human-computer interaction.

The success of Notion's simplified approach to AI raises questions about the future of AI development. Will simplicity and accessibility become the defining characteristics of successful AI applications? Or will more complex and sophisticated models eventually prevail? The answer likely lies in finding a balance between power and usability, ensuring that AI tools are both effective and accessible to a wide range of users.