TravisB
Khoros Alumni (Retired)
Joined 7 years ago
User Profile
User Widgets
Contributions
Re: How we created Maia
Hi Stan, Thanks for commenting. Let me reply to your points with points. 1) I was actually watching when Maia "freaked out" on you, I have no idea how "LOL" triggered the profanity filter, but I think I fixed it. 2) So I like to think of the Atlas Guides as "pre-support" they can't fix your issue, but they can help make sure that it can't be resolved with documentation before pointing you to support. 3)This is a tough question because I don't know your exact situation,ChristopherHashould be able to help you figure out exactly what you need and what we could offer. But to answer for our experience, I would expect to spend about 60-80 hours on Planning and building the initial phase of the bot. I was able to do it with one developer because I kept the integrations to a minimum in the early stages. One thing to note, the execution felt much easier than the conception, figuring out what to offer and how to get humans to that outcome is very time consuming. 4) The "Khoros Bot" is a feature of Modern Chat that is available to our customers. This bot has a really great NLP engine behind it, that I have witnessed in action. Basically you "teach" the bot "intents", these are indicators of what the user is trying to accomplish. Once you have enough intents, the bot will begin using AI to figure out what the user wants based upon phrases you have taught it. Every time someone interacts with the bot, it is learning associations between what they type and what you have taught it. I was watching a chat happen in real time, where Maia opened with the usual menu, the user typed these words "Would it be possible to speak to someone live?" I had not taught Maia this phrase as it is an unusual way of asking this question. There was also not a menu option to speak to someone at this point in the flow, Maia recognized what the user said and stopped the flow, and started a completely different one connecting the user to an agent. basically it received input that did not answer the immediate question, and it read the input and decided to start a different flow. That is AI in action, it was a simple thing, but it was incredible to watch happen. Thanks Stan, Please let me know if you have any other questions or thoughts.
