Looks awesome! Just curious - how much of this is keyword matching vs. actually understanding intent (NLP/ML/DL) in questions/conversations? Also, is there a feedback system that you can use to learn from and improve?
@saip Both keyword search and intent detection are used and yes there are feedback mechanisms to improve.
Report
Interesting. I run a coworking centre (BASE10) in Uppsala Sweden and I built something like this recently internally, using a combo of google drive (for wiki notes) and api.ai to set up a little chat bot (we call the Basebot :-) ) who does exactly this. The one thing I feel its missing though is like all AI, maybe 2/10 times or so from our data it wont know the answer to a question and will give a standard (I don't know answer). This is disproportionately damaging 2 times out of 10 because when it happens the person asking the question loses faith in the bot and then asks a human. Instead I would LOVE to see a feature on niles where if getting a question intent that it didnt know the answer to, a real person (admin) could answer that question (and then save it for future replies). This would give the impression of 100% accuracy and would have people put more reliance in the bot (which saves a lot of resource for admins). If you built that, I would probably swap from my homemade (free) solution to Niles :-)
@jasondainter Thanks for asking Jason, that sounds like a pretty cool homemade solution! What you described is how we think about it too - if Niles doesn't know the answer, you can crowdsource to a person for an answer.
There's a couple ways this is done:
1. If you asked in a direct message, it can ask the question to a channel, even anonymously. Some channels can be intimidating to ask questions in especially for new employees so we found this encourages people to ask questions they otherwise wouldn't have asked.
2. If you asked in a channel, it will post it to the channel so others can answer it. You can also thank others for answering questions since answering can sometimes be a thankless task.
3. Sometimes admins want to add an answer but don't have time right now or need to do some research before answering. In these case they can flag it using a Slack reaction and answer it later.
4. Finally we will automatically detect on channels 1) questions that go unanswered and 2) conversations that should be added to the wiki and suggest admins to take action. One of our mantras is no question left behind.
We think this combines the best of bot helping people and people helping people. We'd love for you to give Niles a try and let us know what you think, you can send me feedback directly at mack@niles.ai. Cheers!
Report
Brilliant idea! Can't wait for this to scale and be adaptable for Airbnb hosts for guest/cleaner communications :-)
@jessiewu44 Thanks Jessie, that's the idea! Often the people with the most answers in a company are the ones that have the least time so this is a way to help them save time.
Report
Interested in this for my team, having trouble finding info on privacy, encryption, and data use - any help?
Report
Is this still around? I’m interested in this for my team but not getting a response from the website.
Thanks for hunting us Kat!
We built Niles to help teams who use Slack to easily share knowledge within the company. I’ve personally felt the pain of creating wikis that nobody used and struggling to find information across multiple tools, and wanted something better.
We designed Niles with a Slack-first mentality so you can create, edit, and manage your wiki directly in Slack. We challenged ourselves to do as much as possible within Slack so you don't have to deal with separate apps and logins. It's AI-enhanced to streamline the process as much as possible and include integrations to other tools (Google Drive, Salesforce, etc) so you can search across your entire knowledge stack.
You can add it to Slack today and we'd love to hear your feedback so we can keep improving Niles. Thanks for checking us out!
FloydHub
Niles
Niles
Niles
Niles
Niles