Page cover image

Overview

Why the Upgrade?

Our new Answer Engine combines our patent-pending Context NLP with Generative AI capabilities. From the start, our conversational AI platform was built to understand and answer hyper-specific questions from your guests. With the launch of GPT 3.5-4, PaLM, and other large language models (LLMs), we are expanding our platform to leverage generative AI features while still keeping control of your brand.

What's the Difference?

  • In our old platform, our Answer Engine used only pre-written responses that you or your staff created and kept updated in the dashboard. In the new system, we can generate the most specific answers by scraping your website and documentation.

  • Your responses become more conversational and granular, as LLMs' main advantage is handling much larger quantities of data from different sources and generating human-like responses.

  • There is no longer a need for your staff to keep all the dashboard responses up-to-date. Ensure your website content is accurate, and use the new scraping tool to update your AI Chat's knowledge volume in a click.

Upgrade Process

  1. To start the upgrade process, reach out to your Account Manager.

  2. Our team will add your web-scraped content and prewritten responses to your new chat. You will have the opportunity to review and publish the responses before testing.

  3. Once you have the new chat link, you can test your chat experience using the instructions below before going live.

Test Your Chat

When you start testing your chat, ask as many questions as possible. We encourage sharing the link with your teammates so that we can gather more feedback.

  1. Open your LLM chat

  2. Type in common questions and review the responses given

When testing your new chat, make sure to ask questions as if you're the end user! Not sure what to ask? Look at your top intents in your Analytics dashboard or escalated conversations in Bridge to see what your users have been asking. Avoid solely testing your chat through buttons and quick replies.

  1. During this upgrade process, use the ๐Ÿ‘๐Ÿ‘Ž on the responses for us to review. For any ๐Ÿ‘Ž please provide a comment with the copy that should be used in its place

When providing a comment on a ๐Ÿ‘Ž response, include the following:

  • Feedback should be as detailed as possible. Consider why the response was inaccurate

  • If the response is missing a button, please add the URL link in the feedback

  • If there is information in the response you'd like to exclude, please note in the feedback

Upgrade FAQs

Do I still have access to my add-on features if I use Context LLM?

Yes! Regardless if you are using generated responses or not, you will still have access to your add-on features.

Do any specific members of my organization need to be involved in this transition process?

We will work with the primary POC and the other members of your organization who have been participating in the meetings. No specific members are needed to work on upgrading your AI Chat. However, you can always invite additional members responsible for managing the content to the upgrading process.

What happens if I don't start the upgrade process?

Your existing chat experience will remain operational, but the portal will no longer accept training requests for the old platform.

What if I have questions about the process?

To speak to an agent, click the '?' icon in the top right corner of your dashboard or by selecting the Click Here For More Support button in the top right corner of this page.

Do you offer office hours for additional upgrade support?

Yes! We host open office hours throughout the week. Tap here to sign up.

How long will the upgrading process take?

It will take approximately 7-21 days. It will depend on how fast you can approve the final version of your upgraded AI Chat.

How do I know where wrong data came from in a generated response?

In the response library, toggle the search bar to "Live Content" and search a keyword word. For example, if a specific slope at a ski resort has been renamed, search the old slope name within the response library so you can identify which response name needs to be updated. This method is very helpful in finding outdated information such as partnerships, names, packages, POIs, and dates.

Click the image to expand!

What if I find that an input is missing a response? How do I add it?

If you ask a question and it's missing a response, you can add it! Either scrape a missing webpage with relevant content, add the information as documented data or the Satisfi Labs team can create a prewritten for you.

How do I view my Thumbs Up/Down results?

Under the Tools > Data section of your dashboard, you will have access to a report called Get Thumbsdown Feedback. This report will not only show you the feedback left by users testing your chat, but also will provide feedback when your users start using your new chat experience!

What do I do when I'm done testing?

When testing is complete, please respond in your LLM upgrade ticket that testing is complete and you're ready to push it live!

Once I complete the upgrade process, does the AI Generative Coverage percentage continue to be updated?

No, this is used to determine your coverage percentage during the upgrade process and is not updated once the process is complete.

Last updated

Was this helpful?