The first killer app of AI for businesses, it appears, is a simple thing: to be able to find information in the morass of files, folders, attachments, incompatible enterprise software apps, and everything else that constitutes modern knowledge work. Notion, which aims to replace most of those things in a single tool, is launching a feature it thinks can help. It’s called Q&A, and CEO Ivan Zhao describes it to me as essentially an all-knowing AI executive assistant that knows everything about everything and can find it in a second or two.
Notion Q&A is available to all Notion users, whether you use it alone or through work, and it’ll cost between $8 and $10 per person per month. The tool has a lot in common with Microsoft’s Copilot and Google’s Duet AI as well as other tools like Dropbox Dash and Google’s NotebookLM. The problem to solve is clear and simple, but everybody’s trying their own way — information retrieval is both a surprisingly complicated AI problem and an increasingly useful one, if you can get it right.
In Notion’s case, Q&A is mostly a mix of search engine and chatbot. If you type in “What’s the office Wi-Fi password,” Q&A will try to find the answer as long as it’s stored in Notion; if you ask “Where are our onboarding templates,” it’ll look for every page about onboarding. Everything Q&A talks about is footnoted to Notion pages, too, in an effort to make sure the tool doesn’t hallucinate and make up information.
Like any good chatbot, the Q&A system can also answer fuzzier questions. During our demo, Zhao pulls up a database someone had made of recent tech articles. He asked for a list of recent articles by David Pierce (that’s me!), which it answered with links to the database pages for those articles. Then, he asked if there were any companies I hadn’t written about recently; the bot, with a combination of Notion data and the so-called “world knowledge” contained within the underlying model itself, said it had been a while since I wrote about Amazon.
That demo is actually a perfect summation of both the strengths and weaknesses of a tool like Q&A. I’ve written about Amazon quite a bit recently, in fact. The LLM knows what a tech company is but isn’t crawling my author page every day; Notion only knows what’s in that Notion database. A tool like Q&A, Copilot, or Duet is only as valuable as the data it can access, and it’s a huge challenge to effectively run your business or life from a single tool. Zhao says Notion is thinking about ways to connect to more data, but acknowledges that it’s a challenge.
One of the other complicated things about Q&A, Zhao tells me, is navigating access and permissions. Your company may all work in a single Notion instance, but everybody has different levels of access to different kinds of stuff, sometimes litigated on a file-by-file basis. How do you make sure everyone can see everything they should and nothing else?
Zhao showed me a demo to explain. He opened up the Q&A pop-up in Notion and typed the question “Who’s on a PIP.” For obvious reasons, not everyone in the company should be able to see the list of everyone on a performance improvement plan, so this was a good test of the access system. When Zhao was logged in as himself, the CEO, it answered the question; when he was logged in with lesser permissions, Q&A just said it couldn’t find anything. That demo went well, but Zhao says there are countless even more complicated examples and ways this could go wrong. Part of what a tool like Q&A can do is retrieve information you didn’t even know existed and wouldn’t find on your own, which makes the guardrails even more important.
Notion’s existing AI tools, which are mostly for generative writing and note-taking, have been popular in the months since they were released. Zhao says Q&A is the beginning of solving the other (and maybe more important) half of the problem: helping people navigate the masses of stuff that come with modern life. It’s not the flashiest use of AI, certainly, to replace files and folders. But for a lot of people, it might be the most useful.