
Dear Notion, here’s your roadmap to catchup with Notis
This article outlines a roadmap for Notion to enhance its AI capabilities, focusing on improvements such as voice input, agent automation, deeper search functions, and better integration with messaging apps. It emphasizes the need for a more cohesive user experience and the development of features like image and video generation, long-term memory, and a unified agent system.
Dear Notion,
I'm glad to see you're finally making good progress on Notion AI. Notion Agents is a promising start—but since it took you three years to get there, we thought we'd help by outlining your product roadmap to catch up with what we've achieved in 12 months.
Here's your detailed plan:
Query Databases, Not Just RAG Search (done, congrats!): It might feel counterintuitive, but it's kind of essential that Notion AI can actually use Notion. What makes Notion so great is the ability to organize and structure information in databases. RAG search is wonderful for finding a needle in the haystack—but try using it to find all the tasks you completed last week.
Voice Input: I think even you can agree that your app is—and will remain—the path of most resistance when it comes to doing anything in Notion. Having Notion AI is a good start, but what users need is voice input. I want to record my tasks and notes while driving my kids to school. Not only that, I want live conversations with Notion AI—brainstorming, retaining context, and getting responses asynchronously after we've spoken.
Agent Automation: While I love Notion automations, it still baffles me that you can't trigger Notion Agents from them. Agents should be an action on their own. Imagine finishing a blog post and automatically having Notion Agent create your social media posts.
Messaging Apps: Let's be honest—everyone hates the Notion mobile app. Make Notion AI available on WhatsApp, iMessage, Telegram, and email.
More Integrations: If we, a company of one, can connect Notion to 800+ apps, so should you. Only 790 to go!
Image and Video Generation: Notion users use Notion to create content. Most content today includes media. Image and video generation should be part of the product.
Deep Search: I'm glad you've introduced web search, but for most tasks—research, blog posts, competitive analysis—it's just weak. We need something beefier to tackle real-life use cases.
Agentic Architecture: C'mon, it's 2026. Do I really need to create my own agents and choose one for every single task? I want to interact with one agent that's smart enough to route to the right sub-agent and adjust its intelligence and verbosity based on task complexity.
Long-Term Memory: While I appreciate the simplicity of storing memories inside Notion documents, there's no consolidated memory layer that all agents can rely upon.
In any case, thank you for being such an inspiration. You've built a fantastic product, community, and vision—and in doing so, motivated us to solve all of this in the past year. Please don't hesitate to steal more features from us. We're happy to help.

Flo is the founder of Mind the Flo, an Agentic Studio specialized into messaging and voice agents.
Related posts
My personal 2025 Wrap
Reflecting on 2025, the author shares the journey of creating Notis from a personal frustration into a successful venture, achieving 80K USD ARR while learning valuable lessons from wins and fails. The personal aspect highlights family life, including the joys and challenges of parenting and marriage, and sets ambitious goals for 2026, including growing Notis to 500K ARR and achieving a balanced lifestyle.
Context Engineering: The AI Skill Nobody Teaches
Discover the importance of context engineering in AI usage. This article explores how providing the right context, including instructions, knowledge, memory, and tools, enhances AI effectiveness, reducing errors and improving productivity in real-world applications.
Choosing the Wrong AI Model Is a Productivity Killer
Choosing the wrong AI model can hinder productivity. Different AI models have unique strengths and weaknesses, and the right fit depends on the task at hand. It's crucial to prioritize speed and efficiency over simply opting for the most advanced model. The Notis approach utilizes a multi-agent system to select the best model for each job, ensuring that quality is prioritized over immediate responses. The future of AI productivity lies in systems that automatically match tasks to the appropriate model, minimizing the need for manual selection.