AI Isn’t a Feature Anymore — It’s the Product
AI-driven automation is reaching a point where the agent becomes the primary way people get work done, turning traditional productivity interfaces into background plumbing that users rarely touch and forcing companies to question whether those legacy surfaces should exist at all.
The Notion Mail shutdown and the sunset of Kotlin Notebook are not routine examples of software product discontinuation; they are a warning shot for anyone still treating AI as a sidecar. Notion is closing its AI-powered Gmail client on September 22 after observing that more than half of its users manage emails without ever opening their inbox. When your own AI agents email management tools make the email client itself irrelevant, keeping that client alive starts to look like a sunk-cost hobby instead of a business. On the developer side, JetBrains is winding down Kotlin Notebook after low adoption and a clear shift in how coders explore and prototype with AI tools. In both cases, AI has moved from accessory to assassin.
Notion Mail: When Your AI Cannibalizes Your Own App
Notion did not kill Notion Mail because the product was broken; it killed it because it worked too well for the agents instead of the humans. The company admits that as its agents became more capable, users handed off email workflows to them, and now "more than half of Notion Mail users manage emails without ever opening their inbox". That is the killer quote: the AI layer became so competent that the GUI became optional.
The practical impact is messy in the short term. The Notion Mail shutdown means the Gmail inbox itself remains intact, but drafts and scheduled emails will vanish if they are not saved first. Custom views, sorting, and reminders built inside Notion Mail will not transfer, and files attached to snippets have to be manually downloaded. HIPAA-covered users must transition off the service earlier, by June 30. Yet Notion’s public line is that this pain is worth it to “go all in” on agents running the inbox. This is AI replacing productivity tools not as a future vision, but as a blunt operational decision: if users refuse to open the app, you stop building the app.
Kotlin Notebook: AI Eats Developer Workflows, Too
JetBrains’ Kotlin Notebook story is less dramatic on the surface, but the logic is the same. The plugin, launched in July 2023, is being unbundled from IntelliJ IDEA starting with version 2026.2 and pushed to the open-source community under Apache 2.0. JetBrains will not ship any compatible version for 2026.3 and later, leaving anyone who cares with what the company openly calls a “compatibility cliff”. That is corporate for “we’re done here.”
The official explanation: Kotlin Notebook “didn’t reach the level of adoption we expected” and AI tools have changed how developers “explore code, prototype, and iterate,” reshaping the workflows that once justified notebook tools. Microsoft made a near-identical move when it deprecated Polyglot Notebooks and pointed developers toward “the next generation of AI-powered coding experiences” instead. The message is blunt: conversational AI and coding agents are the new playground for experimentation, not multi-language notebooks. When agents can summarize, generate, and refactor code on demand, an interactive cell-based interface becomes a nostalgia piece rather than a necessity.
Why Jupyter Thrives While Kotlin Notebook Dies
If AI were destroying all notebook tools, Jupyter should be in trouble. It is not. Repositories containing Jupyter Notebooks grew 75% year over year, from 1.4 million to 2.42 million, and usage nearly doubled within AI-tagged projects to more than 400,000 such repositories. That is not a dying format; that is a format feeding AI’s rise.
The difference is cultural fit. Jupyter was born in the Python and data science world, where exploratory analysis, inline visualizations, and living documents are daily habits. Kotlin Notebook tried to import that culture into application developers’ workflows, but those developers now lean on AI tools embedded directly in their IDEs for quick exploration instead. Google’s Colab underscores the other path: it stays notebook-first, but adds AI agents into the existing workflow instead of using AI as the excuse to exit the category. So AI does not automatically erase every interface; it punishes tools that fail to become the default stage for that AI.
The New Rule: Build for Agents, Not Eyes
These shutdowns expose a hard new rule for product teams: if AI agents become the primary consumers of your app, your real customer is no longer the person staring at the screen. Notion’s experience with AI agents email management shows that when users are happy to outsource their inbox, the interface becomes a maintenance burden rather than a selling point. JetBrains’ retreat from Kotlin Notebook shows the same for exploratory coding: developers have migrated toward AI-native experiences that answer questions and write code instead of providing them with more panels and cells.
The conclusion is uncomfortable but clear. AI replacing productivity tools is not a distant scenario; it is already driving software product discontinuation decisions. Some companies will respond by turning their products into AI-first surfaces, like Colab’s agentic notebook model. Others will quietly sunset any tool that cannot attract humans in an era where agents do the clicking. The winners will be those who design not for maximum engagement time, but for minimum human effort — even when that effort drops to zero.
