Good morning! Is anyone using AI Assistant? If so, what are your thoughts so far? What benefits are you seeing from it, and what issues are you running into?
Great question and interested in hearing feedback from others as well. We have not yet turned it on in our organization.
Is this free or is there a cost?
There's a cost.
Good Morning Jeff, We are using AI Assistant. We love it! We are a global company so many of our offices receive RFPs in other languages. If you say in the additional information - "answer question in Spanish" (or any language Responsive supports) AI Assistant will. We love the "Passive to Active" writing style the most. Our SMEs write in past tense a lot and this writing style is very useful for us in RFxs. The Grammer, Spelling and Punctuation is great as well. Lots of our SMEs have started using it in RFxs because English is not their first language and they are also not writers, they are technical. We do not use the search bar functionality to search for answers, we filter more on our tag. The only draw backs we have seen have been when you choose the writing style Concise or Make Plain, AI Assistant sometimes changes the answer to mean something else. I mean it is doing what it is suppose to do and making the answer more to the point, but you have to read the answer closely because it sometimes takes out important information. But Responsive says we must read all of the answer before accepting AI Assistants recommendation. 😉 Overall, we have found it very helpful and useful!
I echo what Autumn has shared. We have seen essentially the same results on our side. I expect it will improve over time as well, as I think we saw some improvements from the start of the Beta testing to today. We justify some of the cost because our business is strictly prohibited from using public ChatGPT tools, since those feed into a public "database" of information (I am not technical enough to likely explain this properly) and my business provides highly regulated software products. The way this is architected through Responsive allows us to use AI tools with less concern about data issues.
Thank you @Autumn Wenner and @Meridith Hall for this information! We are hoping to learn more about this soon! Do you have any metrics that show the time saved using AI Assistant vs. your prior approach?
We just started using it and there are so many benefits! We had questions and our success manager was able to reply right away. I'll provide the quick FAQ here:
Q: When using the ‘/’ AI prompt, is it pulling content from our library and the public web?
A: The hierarchy for how it pulls is the following: 1) direct Q&A match in Content Library 2) creates content based on context found in Content Library 3) fills in holes from the web-trained learning model.
Q: Is the AI Assistant in Projects and Content Library pulling content from all Business Units and all Collections?
A: It is pulling from all Business Units, but now the Collection and Tag filters will limit what AI pulls from. BU filtering is in progress.
Q: How do we prompt the “First Draft” feature in a Project?
A: In the text field of the unanswered question, press '/' to prompt the AI, then enter your question to generate a first draft. [Note: this is different than what is in the Help Center instructions and demo video.]
Q: When using “Enhance with additional context,” are the prompts dependent on the AI mode?
A: Yes. For example, adding, “limit to 100 words,” will only work in “Concise” mode. You also need to click Concise (or whatever your operator was) a second time to rerun with the enhanced context (the reload arrow will not work).
Q: Will AI be enabled in the Word & Excel plug-ins?
A: Yes! [Note: I did not find instructions in the Help Center for this]
@Cristine Cretan thanks for posting and sharing the discrepancies with the help articles. I shared the feedback with the technical writing team to update the info.
Excellent information, one question, do you know where I can see the costs of using AI? We are RFPIO customers but we do not use AI yet, can someone from RFPIO help us with this? I thank you in advance for the support
Hi @Edwin Diaz,
@travis can follow up with pricing and more information. Cc @Bruce Duffett
Thanks all for leaving excellent questions and Responsive for the answers. I would like to circle back on this Q/A:
... what is a "web-trained learning model?" What actual LLM does Responsive use? GPT 3.5 is only current up to September 2021. GPT 4 is only current up to April 2023.
How can your model be updated every day with the current internet? Models have to be trained to meet Responsible AI once the data-set is determined (internet, etc). That's why they can't just pull from the live internet, correct?
Bing Chat Enterprise was the first tool to access an LLM and then also the live internet.
Can you provide more detail on the LLM and if truly applicable, the live internet that it's accessing?
Hi @Lisa Duplessie,
Thanks for reaching out! To answer your question: when the AI Assistant is using option three (crafting a new response from just the language model), it is not accessing the web directly. It is using the LLM (the AI Assistant is powered by OpenAI 3.5 Turbo) to craft the response. While that LLM is built from publicly available sources on the web, the AI Assistant is using the LLM, not the live internet directly.
Our product team is currently working on upgrading the AI Assistant to GPT4. In our previous attempts, performance was inconsistent (response times and timing out). We are revalidating in our lower environment and hope to roll it out soon.
That said, given the three tiered model we use to tap into your content library, we don't anticipate that it will make a huge difference. Our AI Assistant is designed to work off of our customers' content libraries. However, in instances where it goes to pre trained knowledge (what we call level three), yes, it is possible that GPT4 may yield better results, possibly with knowledge from the live internet.
Hope this answers your questions!
Bruce
