Imagine you could remember anything you’ve written in Workflowy. Even better, what if you could connect the dots across your notes in seconds?
We’ve spent a lot of time (too much?) thinking about how AI could make Workflowy better.
After testing many ideas (thanks to everyone who participated), we’ve focussed on a simple problem:
A note is only valuable if you can find it when you need it.
So, we’ve built something simple: a chat that knows all about your notes.
Use it to:
- Dig up a half-remembered idea.
- Join scattered thoughts.
- Rediscover insights buried in years of Workflowy notes.
We’ve been surprised by the unexpected connections. We think you will be too.
How to Use It
Once the feature is available to you (more on that below):
- Press the ✨ button at the bottom right to open the chat.
- Type a question – it knows about your notes.
- Press ‘Use this page only‘ to focus on just the current page.
How we’ve used it:
- “Summarise what Jesse and I discussed about improving the mobile app”
- “How should I structure a new design system?”
- “Find any notes on product metrics”
- “What have I written about productivity in the last year?”
- “What goals did I set for building the journal feature?”
Open Beta: Free for Everyone
During our Open Beta, this feature is available to everyone, starting with Pro subscribers first, then Free plan users.
You might occasionally run into queues if we get an unexpectedly high usage load, but otherwise, you are free to chat with your notes as much as you’d like.
The feature is not available to everyone immediately; we’re rolling it out gradually to prevent overload. Watch for the in-app message in the coming days, it’ll be hard to miss.
On Privacy
This feature is entirely opt-in.
If you choose to enable it, your notes will be shared with trusted third-party partners (OpenAI, Anthropic, Voyage AI) solely to power the AI features. Your notes will not be used to train their models or for any other purpose. For details, see the Workflowy AI Terms.
Try it, Tell us
What do you like? What’s confusing? What should it do next?
Add a comment below, email us at ai@workflowy.com, or join the Slack Workflowy User Group.
More updates soon.
I would like to be able to input a prompt into workflowy and have it analyze text.
Thank you so much for keeping Workflowy up to date! Not long ago, I switched to another similar software as Workflowy seemed like an abandoned project, only on maintenance but not really improved. It seems now it’s back to development fast mode! Many reasons to celebrate this change!!!
My take on the Wfy AI tool
I am sorry to say that as it is, I don’t see how the tool can be considered functional 😕… I asked it for info in my notes about a person named Léna (being the name of my daughter and thus the person I reference the most in my notes), and he answers that he can find no mention of this name in my notes. Same for fictional characters I am writing about or places I’ve been visiting. When I push him and give him a bit of info, he sometimes ends up finding stuff, but it is all very shallow, when not plain wrong.
So I tried something else, asking about the ongoing projects I am working on, the events I should focus on… It is all clunky, missing a lot of info, or, again, plain wrong.
The cherry on the cake: He often answers that maybe the reason he can not find info about the topic I asked is because it dates from before 2024. But then, when I ask what the deal about 2024 is, he corrects himself by confirming that he has access to all of my notes, regardless of the date…
I don’t know why it does not work, but clearly, it does not 😢
I just tested it out for the first time, and it felt like I could have just asked copilot or a generic AI based on the response I got. There was none of my voice, or my wording or language or terms that I use in the answer AT ALL. In fact it was filled with language that I do not use and the kind of structure and phrases that are not in my workflowy at all.
I have great concerns about AI due to demonstrated dubious ethics of the companies and also the negative environmental effects of AI’s high energy requirements. I also just plain don’t trust those companies. If there is a button in the corner, I can guarantee that at some point my fat finger is accidentally going to hit it. If this happens does my data immediately get sent to an AI company? How can I make sure this doesn’t happen? Thank you.
PS Workflowy was already great as is.
Love it. One thing I found peculiar: when I ask it “what should I focus on?” it includes items I’ve marked as done/completed (I don’t want to focus on things I’ve marked done). I would prefer that it leaves those out. Thoughts on implementing that or alternative strategies? Thanks!
Another thing: I asked it to tell me what I know about AI and it asked me if I wanted it to reference my notes. By default, even if my question is imprecise, it should assume that answers to my questions are based on my notes.
Hey there, I have question about the following function:
–> “Press ‘Use this page only‘ to focus on just the current page.”
Does this mean, Workflowy only sends this page (and all bullets below) to the AI partners and nothing else from Workflowy?
The reason I ask: With some projects, I’m legally forbidden to use AI, with others I am not. So I have to make sure that none of the data from the “forbidden” projects are shared with AI companies.
This is horrible! If I cannot turn this feature off forever, so that I never even have to look at another AI icon again, I will find a new program! It would be a pity because I love workflowy, but I cannot stand AI and I can find something else that will work.
It’s not worth being constantly harassed by “why don’t you use our plaigarism machine?” I have enough stress in my life already.
Hey C W, you can definitely opt to not enable the feature and that would be the end of it.
Long time pro subscriber here, thinking of canceling my subscription.
How much access have you provided AI companies to your database ? Now and going forward ?
The writings of your users are hugely valuable for AI training purposes.
Especially as the quality of the data that those companies are able to scrape (illegally) off the web is diminishing.
I am waiting for a clear statement of Workflowy how private the data is that we have been entrusting you with for all these years (12+ for me).
Hey Patrick, we have not provided access to any data to any AI companies without user consent.
What we have done is added a feature that is completely opt-in. If a user does not opt-in, no data will ever be sent to any AI company. Period. Also, none of our providers are allowed to train their systems or models on any data we send to them.
Hi Roldofo, thank you for the clarification. I hope Workflowy will remain cautious in balancing new features and privacy.
I tried to read up on Workflowy’s privacy policy but ran into a wall of legalese. It would be great if we can get that information in human-readable language.
There is quite literally no reason to trust any AI company, whose business are built on mass plagiarism, to simply not abuse data sent their way. Stop seeing dollar signs and start seeing reason.
“It already knows your notes!” That one sentence sets off a privacy alarm. I haven’t tried this — I can see how useful it would be, but allowing this AI to data mine my notes is another, and you are using a third party to operate this service… all sounds a bit dodgy. Shouldn’t you prove to us where and how this data is used/collected. And there needs to be a opt-out options — for good, or temporary….
Hey david, thanks for the feedback. Apologies for alarming you with the wording we used. We are definitely going to update how we describe the feature and make it really clear what’s going on if you decide to opt in.
By third party we mean any company that isn’t Workflowy. We don’t host AI models so we use third parties Openai and Anthopic. Every company that uses any sort of AI in their product or service that wants to use Chatgpt or Claude does this.
But essentially, none of your data is sent to any AI company if you decide to not opt in, period.
Also, none of the companies we work with are allowed to use your data to train or improve their models, or do anything else with your data.
We send them data to process, they send us the response and then your data is either 1. Never stored (in the case of Voyage AI), or 2. Deleted after a maximum of 30 days (this is done by both Openai and Anthropic in order to prevent abuse of their systems).
Glad I was able to make the AI features disappear by toggling it on and off again in the settings. After 13 years with Workflowy in my daily life I definitely use it to help me think, but generative AI would honestly just sabotage this; even a floating sparkly in the corner just feels like an interruption. Here’s to hoping Workflowy keeps just working perfectly instead of deteriorating because of forcing AI.
Hey Jason, thanks for the feedback. From the moment we decided to offer AI in the app we knew we wanted to make it 100% opt-in. If you don’t want to use AI in Workflowy, you can turn it off and you won’t see it in the interface.
Amazing news! Was about time! 🙂 Looking forward to its release. We are already using Workflowy beta version to see if we get it soon!!! Thanks!
Hey John, thanks for the feedback. Hope to get it into more user’s hands soon.
Wondering about the impact this would have on the cost. I like Workflowy the way it is and I have a sub until August. Please don’t tell me you’ll randomly up the subscription fee. If you do (which is the logical thing) I’d really love if you maintained a no-AI tier with all the existing functionality.
Hey Tao, thanks for your feedback. We’re still figuring out what the pricing is going to be so take what I’m about to say with a grain of salt. But I believe it’s very likely we won’t change the existing pricing and instead offer AI as an add-on or a different tier for the users that want it.
Wow, this feature certainly seems to be stirring up some strong opinions! I have to say, I was a bit surprised by some of the reactions – didn’t realize AI could be such polarizing hot topic.
Personally, I’m really excited about this addition. That said, I am getting more skeptical about the practical, everyday uses for many of these chat bots, so it’ll be interesting to see how this one pans out.
On that note, could we get a little more insight into the tech behind it? For instance, is it using a RAG model?
It’s clear you’ve built up a dedicated community (myself included!), and I have a feeling many of those who are hesitant now might just come around once they see it in action.
Hey Robertpaulson13, thanks for the feedback. We’re working to give our users access to one of the fastest growing technologies while giving them the ultimate say on whether they want to use it or not. That’s why the feature is completely opt-in. We understand not everyone is onboard with the tech and we respect their choice.
With regards to the technical side of things, I think at some point, possibly when we do the final release of the feature – we’ll include more details about how it works in the background for anyone interested. It might also help some folks better understand how the data is used and
Hey Mariano, we’re still rolling out the feature – please sit tight!
I’m not supposed to share work content with outside AIs–we’re only supposed to trust the Microsoft AI. Also, honestly I don’t want to apply AI to my notes. It would be great if settings could have an “AI off” switch to opt out of even seeing the AI button, that way I can assure IT that I’m not risking sharing data with an unapproved AI.
One of the things I love about Workflowy is how simple and unadorned it is for the most part–it doesn’t email me, or clutter my screen with loads of functions I don’t want, or hide useful info. I know you have to continue to innovate and improve, but as far as I’m concerned you’re pretty much perfect and I don’t want to lose that 🙂
Hey Victoria, we hear you and thanks for the kind words! You’ll be able to completely hide the AI options from the settings panel if you choose to opt out. None of your data will be send if you don’t opt in.
Wow – I just closed my WF account of many years to use a local only app – this might have changed my mind.
Hey Dan, lots more to come.
Stoked to hear about the “Chat with your notes” feature! Great Job!
New Request:
Recently, I’v been thinking about how powerful it would be if WorkFlowy could function more like a browser, not just a note-taking app.
What I’m dreaming of:
I recently started using Arc Browser, and while it’s amazing, I keep wishing it had WorkFlowy’s outline-based structure as its foundation. Or vice versa—that WorkFlowy could become a full web browser.
Imagine:
Each bullet (node) could be a live browser tab, a link to a webview,
• Nest bullets to create tab groups/workspaces
• Zoom into bullets to switch contexts
• Notes + links + tabs all organized spatially in the outline
• Tabs persist across sessions naturally via WorkFlowy’s structure
This would create a seamless thinking + browsing experience—a mind-mapped web, rather than flat tab bars.
Why this matters:
• WorkFlowy already provides perfect cognitive scaffolding for thinking and organizing.
• Adding basic browser functionality would make it the ideal tool for research, learning, and creative workflows.
• It could start as a desktop shell (e.g., Electron-based) that uses WorkFlowy as the OS for browsing.
• It could become its own Chromium-based browser.
Is something like this on the radar?
Hey Roarke, that’s a very interesting idea. I don’t think we’ve really thought about that but I’ll share it with the team.
This is a very serious update, thanks for your work!
It would be great to have the an option to have an AI attached to WorkFlowy, as a handy chat with options to ask questions
and copy answers, but not inside WorkFlowy, with an access to the notes content.
Some kind of this model is a Flow Us AI (or Buildin).
Hey Alexander thanks for the feedback. Just to make sure I understand, do you mean a standalone AI that is connected to Workflowy?
Yes, Rodolfo, something like this.
Hi, I have been using your app for years now, as a writer and artist. I have thousands of notes written in it, and it has genuinely helped me for organizing all my work. I have spent over 10 years writing and building my skills as a writer, and Workflowy has been a crucial tool in organizing information I otherwise would have lost. Throughout all these years I have loyally praised Workflowy to my peers for its simplicity, versatility, and robustness.
The LLMs created by companies such as OpenAI, Anthropic, and Voyage AI, are built on datasets entirely made up of stolen copyrighted works from across the web, including from writers and artists like myself. These AI models do absolutely nothing to help me or my work. They give widely inaccurate and unverifiable information, and are typically quite poor at conducting the tasks they are commanded to. Furthermore, the electricity-costs are unjustifiably large for what are, on a technical scale, and for all intents and purposes, giant plagiarism machines for people who are tired of thinking for themselves.
I will be actively moving all my notes to a different program and telling everyone I know and more to NOT use your app, until you roll back everything having to do with “AI” and never touch it ever again. If you completely drop it, and shift your stance against AI, I’ll move back and become a Pro-Plan user. Sincerely, a gravely disappointed user.
Thank you for putting all of this into words! I am a long-time subscriber, and I’ve always been an avid and vocal supporter of Workflowy—and I still certainly want to be—but implementation of AI (and its plagiarism) makes me worry about the security of my data. AI just makes the world worse, not better.
I am glad I am not the only one who agrees! Opt-in or not, the use of these third parties for AI implementation is not ok. They are entirely built on the use of the intellectual property of creators and thinkers from around the globe without their consent. I was hoping workflowy would be better than this, but unfortunately most companies only see dollar signs when they see AI, and it seems like workflowy wants to be like everyone else.
I think they need to focus on features that relate to workflowy’s simplistic and easy-to-use form factor that made it special in the first place. The notes are there to help me think, not think for me.
While you have some info you might not be as informed about how these things work as you might think. They are in the most literal sense not plagiarism machines… They don’t store data, at all, that is how all the current search tools you have probably been using for years work via what is called natural language processing (NLP). Models these days are now trained primarily with synthetic data (data the llms make up themselves) anyway and the big foundational models even pay for other datasets these days.
This is most likely what is called a RAG tool. Your content is stored in a mathematical representation in a several hundred dimensional matrix data point (embedding vector). When you ask a question, some math happens to find your notes that are related and then those are synthesized and shipped off to a service like you mentioned to summarize.
If you have ever used any google or microsoft product, any social media tool, almost any website with search or products, it is likely everything you have seen, written, clicked, typed, or even hovered your mouse on is tracked somewhere. Indexed and for sale. While this is not to say this reality is not terrible, I would be remiss if I did not call out the selective outrage you are deploying against a project you otherwise love.
There is no doubt that data privacy is important and we should advocate to protect our own content. At the same time, abandoning the tools and flows we have used for decades because of AI rises to **shakes fist at clouds** behavior.
You seem to have selectively misunderstood the whole basis of my comment. My criticisms of AI as a whole are entirely on the third party companies, who do in fact steal data and copywritten works in order to train their AIs, that which will then regurgitate that data to users inaccurately and without credit or verifiable sources. That is literally plagarism. Coupling that with the insurmountable electricity cost these third party companies demand to power AI, I find the workflowy developers’ choice to use their resources to power this new feature to be completely unacceptable.
Neither myself or any self respecting writer wants their work scraped and used for AI, inside or outside of workflowy. I am not interested in relegating my thinking to a third party program that doesnt even work correctly half the time in exchange for more smog in the air and less water to drink, and I will condem anyone or anything that enables this practice, including workflowy.
I LOVE this and have been sorely hoping you’d infuse AI into Workflowy soon. To hopefully assuage some concerns with personal data “leaking” and being used to train models, I’d love to see a strongly worded security statement about how my data is not visible to anyone, used to train the models used, etc. Great job! Can’t wait to try it out!
Hey Braden, thanks for the feedback. We could definitely be clearer with the specifics of data handling and privacy in out AI terms.
As Julie and other Mark have said, thank you for making this opt-in. I have no interest in this feature and don’t want my use of WorkFlowy to even unintentionally require the resources and risks of AI (a la Google jamming it into search results). I, for one, do not trust the “trusted partners.” I truly love WorkFlowy. Please don’t become an AI company.
Hey Mark, thanks for the feedback. We know not everyone wants AI features in their products – that’s why we made everything opt-in. We’re also working to bring more core features to Workflowy soon.
I think this is fantastic news. I use AI tools every day and have found myself using Workflowy less because of it. It’s time for me to re-up my Pro plan with Workflowy, and AI integration was going to be the dealbreaker. It couldn’t have come at a better time.
We’re working hard to bring more core Workflowy and AI powered features features soon!
Absolutely opposed to AI coming anywhere near my Workflowy—I am so relieved to see that the product will be opt-in rather than forced upon us! Thank you!! I think your existing search features work perfectly well!
Hey Julie, thanks for the feedback. We know not everyone wants AI near their data and that’s completely reasonable.
Sounds interesting, looking forward to trying it out!
A few weeks ago I also thought about AI and Workflowy. It would be cool to habe the ability to use ChatGPT (or others, through API) directly in Workflowy and then have the answer as sub bullets to the question. Often I use it for research and copy it after into Workflowy. Having it there directly after searching and being able to edit/delete as I wish would be really cool.
Thanks for the feedback Adrian, we’re trying to roll it out to everyone quickly so hope to hear what you think after you’ve tried it out.
You’ve detailed “chat with your notes”: but notes in app are clearly denoted as below bullets, so yr stating it ONLY searches notes (which makes little sense), so what is it? Maybe edit yr post?
Hey morgs, thanks for the feedback. I can see how it might come across as confusing since we also call the text under bullets ‘notes’.
I had the same confusion, so I looked through the existing comments since there was a pretty good chance that it wasn’t just me. Referring to regular item text as “notes” is perfectly natural for so much of what Workflowy is used for, and a web search just now showed me that this isn’t the first time that this has been done. I’m thinking that text-under-bullets needs a new name.
Awesome! Does the feature support any other language exclude English?
The feature supports many languages other than english but there is currently no way to set the default language – I’ll share that with the team.
Genial !!! Felicitaciones al equipo y la decisión tomada me gusta mucho. Espero a poder utilizarla prontamente. Abrazos
Muchas gracias Carlos, no tendrás que esperar mucho.
Gracias Rodolfo
Just a question, when you tell the AI to “Use this page only”, will it drill down all the nodes, whether they’re expanded or not? or will it look only at what’s literally visible on the page?
Hey Jim, that’s right. The AI will use everything at that zoom level whether it’s visible or not.
how’s the pricing when it’s released?
Hey John, it’s something we’re still figuring out. The open beta will give us a better idea of what the average use will be and that’ll help us price it appropriately.
If we can use an API key to reduce cost/load on WF, that would be a great option.
Love the idea, but not comfortable sharing my data with a 3rd party — not sure if there’s any way around it. But that’s a dealbreaker for me.
Totally understand Mark – thanks for the feedback.
Awesome! I can’t wait to see the magic button! Thanks!!
Hey Rafael, we’re rolling out as soon as we can. Can’t wait to let you try it out and kick the tires.
Congrats. Nice to have it in Workflowy and to have it as an optional feature.
An idea: would it be feasible to give AI limited access?
For example, I could have a “Work” bullet that I’m interested in use with AI but a “Family” bullet that I’m not interested in AI having access to.
A way to lock some bullets o to allow only some bullets would be ideal to not force the user to opt into a dilemma of a YES or NO in their accounts.
That’s an interesting idea ronsu – thanks for the feedback.
🤍
It’s a must.