For enterprises, the first killer application of artificial intelligence seems to be a simple task: being able to find information in documents, folders, attachments, incompatible enterprise software applications, and everything else that constitutes modern knowledge work. Notion is a product designed to replace most of these things with a single tool, and it is launching a feature it believes can help users. It's called Q&A, and CEO Ivan Zhao described it to me as an omniscient AI assistant that knows everything about anything and can find it within a second or two.

Notion Q&A is available to all Notion users, whether you use it individually or for work, and it is priced between $8 and $10 per person per month. This tool has many similarities with Microsoft's Copilot, Google's Duet AI, and other tools like Dropbox Dash and Google's NotebookLM. The problem to be solved is clear and simple, but everyone is trying to do it in their own way - information retrieval is both a surprisingly complex AI problem and an increasingly useful one if done right.

In the case of Notion, Q&A is primarily a mix of a search engine and a chatbot. If you type "What is the Wi-Fi password for the office," Q&A will try to find the answer as long as it is stored in Notion; if you ask "Where is our onboarding template," it will search for all pages related to onboarding. All the content discussed by Q&A also has footnotes to Notion pages to ensure that the tool does not generate illusions or fictional information.

Like any good chatbot, the Q&A system can also answer some vague questions. In our demo, Zhao opened a database of recent tech articles made by individuals. He asked a question about a list of David Pierce's recent articles, and it answered with links to the database pages of those articles. Then, he asked if I had written about any companies recently; the bot, by combining Notion's data and the so-called "world knowledge" contained in the underlying model itself, said that I hadn't written about Amazon for a while.

This demo was actually a perfect summary, showcasing both the advantages of tools like Q&A and their weaknesses. In fact, I have written a lot about Amazon recently. LLM knows what a tech company is, but it doesn't crawl my author page every day; Notion only knows what is in its Notion database. The value of tools like Q&A, Copilot, or Duet depends solely on the data they can access, and effectively running your business or life with a single tool is a huge challenge. Zhao said Notion is considering how to connect more data, but he admits it is a challenge.

Zhao told me that another complex aspect of Q&A is managing access and permissions. Your company may be working in a single Notion instance, but everyone has different levels of access to different types of things, sometimes even on a per-file basis. How do you ensure that everyone sees what they should see and not something else?

Zhao showed me a demo to explain. He opened the Q&A pop-up in Notion and typed the question "Who is on PIP?" Obviously, not everyone should be able to see the list of personnel on a performance improvement plan, so this is a good test for the access system. When Zhao logged in as himself, the CEO, it answered the question; when he logged in with lower permissions, Q&A simply said it couldn't find anything. The demo went smoothly, but Zhao said there are countless more complex examples and ways things could go wrong. Tools like Q&A can do part of the job of retrieving information that you don't even know exists and can't find yourself, making security even more important.

Notion's existing AI tools, mainly for generative writing and note-taking, have become popular within a few months since their release. Zhao said Q&A is the beginning of addressing the other half (perhaps more important) problem: helping people navigate the vast amount of things in modern life. This is certainly not the most eye-catching use of AI, replacing documents and folders. But for many people, it may be the most useful.