So first off, I have paid over $10k working with a developer to build something similar to this and when I found this - I fired him.

It's a new product so there is room to grow, however on the initial set up. I have a fully built and developed chatbot that I was trying to create on my own backend with much difficulty and cost within one day.

It works today for anything you could possibly use it for, but it also has a lot of potential. Including connections to AWS servers, airtables, and other components to really supercharge your models.

I like that you can optimize each bot with specific models like OpenAI/Anthropic/Llama/Etc to really tweak the optimal outputs you're going for. This deal is a really good one and I got over 15k credits to start working and burned through about 1,000 training my bot so I've got a lot of room to grow with the current plan.

My only initial comments are that the AI Suggestions are a little slow to populate, but very good questions when they do populate. It cost credits to get those suggestions which is not controllable (you can either turn it on or off, not decide when the suggestions show up). Because of this, it causes unnecessary token burn (4 tokens with output w/ suggestions vs 1 token without suggestions) - even in the training model. This is obviously because any API call is going to cost IC money to do that, but I did feel like users would benefit from a little more cost efficiency in their training and development.

It would be possible to mitigate this with further development of the integrations so I didn't allow it to impact my review scoring.

The voice chat, unfortunately is an all or nothing proposition which essentially makes it not useable for me. The idea of a voice only dialogue with AI bots is still a ways away unless you're using ChatGPT which detects interruptions, etc. To make this more UX friendly, I would maybe allow a user toggle within the chat to turn "Voice Response" on. This way users can type in or use buttons/cards to advance the conversation and get an audible response so they don't have to think too much and keep the conversation going.

The response time is manageable with some knowledge base items. I found I could speed it up by creating more streamlined summary documents to guide it compared to the full length content, however I lose quality in that respect which isn't preferred. Maybe a slider like the creativity one to help determine what the bot should focus on, quality or speed.

I couldn't figure out how to do the "Whitelabel" or custom domains. I think the white label was, obviously, the way to rebrand these and sell them on behalf of IC under DBA branding, but it skipped the custom domains. The absolute last thing I want on my bots is anyone elses brand or domain. When the URL app.insertchat comes up as well, it seems kind of suspect so the sooner users can build custom subdomains (SPECIFIC to the bot) the better.

It would also be helpful to get more guidance on how to build within the ecosystem. For example, the starter prompt for the AI bot (Sales, etc) are ok but I would write them way differently and in fact, did. I could only do this because I am already very familiar with prompting and frameworks, but many users might not be. More videos/wiki/how-to/explainers/tool-tips would be helpful.

I think ChatGPT failed big time by not allowing users to create storefronts, and instead went with a marketplace where only popular bots are seen and because of the algorithms are the only ones ever seen. Showcase user bots, highlight people who build cool things and show them to the audience, it will go far for grass roots growth.

Finally, these bots are made for interactivity but I feel like they sold themselves short not designing them as content engines as well. This would not only increase the user time on system, it would also increase token spend because a chat convo is a lot different from sitting down to build a webinar outline and content.