Fini AI Reviews

1. Sign up page doesn't work properly. Had to try various clean browsers to get redirected to 3rd party authentication.
2. 1st bot setup failed with an error "Oops! We encountered an issue during adding links. We apologize for the inconvenience. " after hours of waiting. Could not delete or add another bot during this time so all I could do was keep waiting and checking - very bad first impression.
3. No indication of whether the bot is operational or in whatever state. The 1st failed bot was not found and I was not sure what to do next. The documention stops at configurations and does not explain what to expect next.
3. One of the 2 links failed. The bot worked after removing the 2nd link. But the bot still takes information from the 2nd link for answers.
4. Tried 2 bots, both have problem with accuracy, providing answers not from the supplied links or referencing wrong part of the information. There are cases where the bots provide conflicting answers to questions with exactly the same words.
4. Doesn't seem to support double byte characters for bot names. The UI accepts the inputs but does not list the names in either Setup or Ask Fini (chatbot).
5. Some times responds questions in configured language PLUS the same answer in English .

The bots are simply not useable with this quality.

- 1 taco for quality of answers, since this is what Fini X claims to be better than others.
- 1 taco for the incomplete and confusing UI implementation.

Fini AI is the most mind blowing product I bought from Appsumo tho,I just leave a short review, before fully utilizing Fini AI in our use.

Standalone Link:
* Integrated Fini AI's "Standalone link" with "Gobrunch", virtual office platform, for internal use.

Base Prompt:
* Fini AI's "base prompt" is highly versatile, initially aimed at customer service but adaptable for wider applications. There’s a learning curve as it requires custom development, similar to ChatGPT's "custom instructions". and "baseprompt" is way better than that of "Re:tune" that I was bringing my GPT4API.

Knowlege base only: I'm trying to make baseprompt to let it work like "Dante AI" right now, that stick with trained dataset more. I have to say default prompt by FiniAI is good, but I just customize baseprompt and "knowlege only" setup won't limit to dataset trained, we need to make better baseprompts.

Chrome Extension:
* Room for improvement as it currently supports single-query responses. Enhanced functionality achieved when linking a tailored "base prompt" for chat interactions, outperforming some features of Voila, like translation or summarization.

Security Measures:
* Compliance with ISO, SOC2, and GDPR standards sets Fini AI apart from many chatbot solutions, emphasizing a strong security framework.
Deployment:
* Appreciate the ease of native integrations and the continual product development.
Answer Quality and Speed:
* Faster response times compared to Dante AI. Answer quality is good, though dependent on the "base prompt" setup.
Language Quality:
* Noticed superior language quality in Japanese and Korean responses from Fini AI compared to ChatGPT-4 by OpenAI. super superior to any others I have used , very promising for customer service use.

Look, I'm really pissed off about this trash deal.

Starting with the forced third party login, that can't even be changed to the fact that their offering is simply a scam.

They don't offer GPt-4 quality outputs no matter how they want to cut it with their FULL OF LIES answer and absolute cop out of MULTI-LLM Layer.

It's absolute BS.

I've conducted tests on the model and have verified that it is not capable of GPT-4 level output. And even sometimes GPT-3.5 is superior.

I will not explain in detail the kinds of prompts because I don't want them to make it even harder for other to discover how limited it is.

But simply put, the quality is simply not there, this is not GPT-4 and actually inferior to GPT-3.5 Turbo.

So I picked up the highest tier knowing it will become a good investment. Hence, I've knocked off 1 star for now - waiting on their roadmap to come to fruition.

These guys are Y-Comb grads for a reasons. If you read their docs you'll know why. They are deeply focused on security and enterprise requirements, which means you can count on it being top notch.

The product is not quite there yet, but is still decent as is. But this is only the beginning and there's so much on the roadmap. As sumo-lings we often have an opportunity at this early stage to push forward ideas, which ultimately can shape the product.

In closing, they're well funded and I'm looking forward to reaping the rewards of this small investment.

Don't stop building team!!

Let me start by saying I purchased Tier 3 which is the highest tier available at AppSumo

TLDR; this product has many flaws and is not ready for production. There are better customer support AI available on AppSumo that performs better than Fini.

The bad:

1. There is no way to prevent the bot from answering questions irrelevant to the documentation provided. To me this is the deal breaker because people can exploit and use the bot as generic GPT bot. Apparently, in order to do this we should purchase the custom tier (starts at $299/m) to get access to the inhouse built models.
2. They advertised that Tier 2 & 3 users are using GPT-4, but my Tier 3 account is apparently on GPT-3 (I asked the bot what LLM they are using and they replied with GPT-3)
3. There are a lot of bugs that needs to be ironed out, for example when we're using AskFini section to try out the bot, sometimes the instruction prompt are leaking when switching from one bot to another. I also cannot make the chrome extension work (setup the API and instruction, but no chat section is opening up after that)

The good: Can still be used as generic GPT bot with static instruction prompt.