Slite's new AI assistant – Ask – answers questions based on your team's documentation. Take the pain out of finding info, learning from the past, and getting new team mates up to speed. Have a question? Just ask.
I'm so happy to unveil Ask to the world and hear feedback, finally!
If you're curious on the genesis and reasoning behind Ask, read on, else thanks for your support and make sure to join the waitlist!
Large Language Models are not recent, we had been exploring them for a while, but they were brought to a whole new level with OpenAI and @samaltman6 work on chatGPT a couple months ago.
The obvious application in document tools was be to bring it *inside* the editor, to generate content.
That's what various tools, such as Notion and Craft did. We developed this internally as well, and might still release it in some way, but something felt off, or not good enough compared to the magic of chatGPT.
Since the beginning, Slite has been a tool for teams. Not a no code tool, not a 2nd brain for individuals. We optimise it to centralise all your team information, wikis, projects, notes, key decisions, and not for the sake of it.
The goal is of course to deliver this information back. And looking at chatGPT, we knew the future of our tool needed to go in a similar direction.
It was not easy, the team worked extremely hard on this, put a lot of intelligence in the solution, and managed to pull through a great first testable version!
Join the waitlist, we'll be gradually rolling it out, starting with Slite users and teams that see a very strong interest in their business, let us know when you join the waitlist!
We're far from done, we have a ton to learn and to refine on.
The first next brick, for instance, will be to help users deprecate content, and do so automatically. That's the limit of GPT for your docs: there is no page rank like on the web, saying what's true or not, what's still relevant.
We need to make it easy for Slite users, so that Ask can answer based on sources they trust.
Thanks all for the support, thanks so much @tiboel for hunting us, share and send your feedback!
The team will be there to answer all day long.
Hey there,
I'm beyond thrilled to share Ask Slite with the world today!
For the last few months, we've seen players trying to bring the power of advanced language models like GPT to our documents.
But all we've seen so far has been about generating content. Writing more has never really been the issue. On the contrary.
The future is not in creating more noise in teams, but unlocking the vast amount of info your team put years to bring together.
And with Ask, the Slite team finally cracked it!
I've been using it for the past week and I can honestly say that this will change the way we work and collaborate as a team.
The Slite team has worked incredibly hard over the past years to build a flawless product for all team documentation. With Ask, it now combined it with the power of large language models, and it's brilliant.
It's like having a team assistant you can query on anything your team has ever thought, or done.
The product is a game changer and will be very helpful in setting up a knowledge base quickly. ✨
How long has the team been working on it?
Best of luck to the entire team on the launch. It will be a #1 product for the day. ⚡️⚡️
@worklab The team has been working on it incredibly fast and hard for the past month and a half. It's just the very first milestone, there is such a big potential to unlocking after! Stay tuned ⚡️⚡️⚡️
Congratulations on your launch @christophepas!
This sounds interesting! How does Ask ingest the documentation created by my team exactly so that it can give me pin-pointed answers?
@christophepas@ug911 Great question!
Ask generates a result by performing a semantic search across all the documents you have access to in your workspace. When a user types a question, it searches for docs (or docs fragments) that are closest to that search query. It refines even further by reranking the results based on multiple criteria like how recent the documents are and their status to get a relevancy scores. Thus, the answer can come from a single source or be synthesized from many. The top results are then fed to GPT (Generative Pre-trained Transformer) as a prompt, which then completes the text with an answer.
You can try it out on slite.com/ask, and share away ⚡️
I've been using Slite for years to document decisions and learnings.
Ask is a major step forward in how we work and collaborate, beautifully integrated into the product we already love.
@christophepas and team, you nailed it once again. Bravo, keep it up 👏👏
Can't even begin to say how impressed I've consistently been by the team at Slite. They are the Apple of the team documentation and collaboration space. They actually take a moment, see what the new tech and features are all about, and then implement it in the most thoughtful and impressive way (first Discussions, and now Ask).
Everyone in SaaS knew they should implement GPT, and that they did – the same uninspiring me too way. Autogenerated content, manually highlight chunks of content to summarize, blah blah blah. Then Slite asked "why?" and Ask by Slite was born.
What does Slite know that others seem to miss? Whelp, we have SO MUCH DAMN CONTENT. Internal company docs, customer docs, coding reference docs, content idea docs. We then share some of these docs with customers, and others with the public, so now our now content is theirs. (Yeah... you can even make a white labeled website around your Slite docs) – moral of the story is: everyone already has the content, too much content!
Any company using a documentation and collaborative note taking tool really just needs one key feature. A way to find what you're actually looking for. It's obvious, no?
But search was obvious, until we realized with Ask that it wasn't good enough. That you should be able to get a result that combines information from numerous documents, all that disconnected information now connected to give a simple answer to the question you asked.
We have the information, now Slite has the Ask. I'll say it again, it's obvious – well, Slite made it obvious 👏
Thank you @christophepas & team for continuing to do what Slite does best – taking the time to ask "why?" before building.
@christophepas@alexhbass ❤️ thanks man, you've always been here to challenge us and give feedback! We couldn't be more grateful! Having users like you along the way makes the ride even more vibrant!
bold move, really like the idea, we did it ourselves internally and it has been a game changer for internal process and support teams around the world. Quite interested in any future development! Where do you see this product going in the next few years?
@ant0ine_gt love to see a fellow "content overload knight" and would be pumped to hear your learnings from building and implementing this internally 😻
As for the next few years,...🥁 @christophepas's vision is bright so I'll let him share it 🤩
@ant0ine_gt Great question!
It goes beyond static knowledge, and can really be a work assistant. Our goal is to better understand your content relevancy, the timeline of your docs, to then tell you what happened, when, who does what, and have answers way more accurate + based on the events that really happened.
The wiki/handbook use cases are the top of the iceberg!
@ant0ine_gt we're also looking into ways to automatically summarize pending decision thread to re engage participants (currently trying it).
If you have any ideas we're up for it!
Wow, very interesting and helpful. This product can be used to store help articles and integrate them into help desk software and then use it to automatically respond to customer inquiries in the support (just wanted to share this idea). Good luck!
Wow 🤯, I'm so excited to see Slite's new Ask feature! 🤩 It sounds like an incredible way to make sure teams have access to the most accurate, up-to-date information. Can't wait to check it out! 🤗
Love the shared excitement, @paolo_degiglio 😻 I wish I had Ask every time I joined a new team so I could get instant answers... can you imagine your first week at a new team with an AI assistant that knows it all? 💫
Congrats on the launch team, this is an absolute game-changer!
That kind of feature where we say “that is so obvious” now we have it in front of us.
Big kudos to the team, I can totally see this becoming a new standard! 👏👏👏
@robi Thanks Sebastien, that's what we're hoping for! Reducing the burden ofcontent consumption, and opening the door to a next generation of knowledge base ✨
Brilliant! Spent way too much time on getting the right info in the middle of a doc, or looking at the outdated doc.
Glad Ask by Slite will help us all at Fairmint save time and get the right info in seconds. 🚀🚀🚀
This would be plug and play for any sort of internal support requests/questions (HR/on-boarding/etc) and could be used to help relieve pressure from external support requests. I see it as complimentary to existing NLP-based support chatbots.
It's definitely the next evolution of support automation, whereas prior the bar was set at efficient indexing of knowledge base articles or otherwise.
What's on the roadmap going forward?
@cjries Super interesting to hear! I't amazing for support -> check it out on our help center, you can actually try it! https://slite.slite.page/p/tQruG...
But it's absolutely game changer for team work too.
The big part of the roadmap for Ask comes to improve the results for all the part of the documentation that is not that *static*. If something changes often, you'll have many variation of the "truth" -> we have the tool to detect which one is really relevant and give you a great answer, even on these kind of questions.
Also, answer all your question on your teammate, on recent activities etc would be incredible. Stay tuned!
Great idea! Having Ask available to answer questions about team documentation quickly and conversationally has to be a lifesaver for companies. What types of questions does the AI understand?
@codecamcode 100%! It can answer many questions based on the documentation in your workspace. We see that it's especially valuable for basic company information, handbooks, policies, culture-related things, well-documented workflows, onboarding 🙌
Been playing around with this and its fantastic!
It's mind blowing how now having a strong written culture can create a super helpful knowledge base that gets you the answer to any basic question you have.
Such an incredible idea! This solves a real problem we all face and it's probably one of the best ways to use AI in our work tools. Congrats team and best of luck with the launch!
Slite