Skip to main content

AI Support Chats

Here is how to setup and train your AI support chats for your volunteers

Ash avatar
Written by Ash
Updated over 2 weeks ago

Please note: While this feature is being tested, it's only available to a limited number of organisations. You can apply for your organisation if you're interested in helping Volunteero test this feature and validating the value for your organisation. There are limited number of spots available, we can't guarantee your organisation will be able to join the closed beta. Volunteero app version 2.30.0 has support for this beta.

Overview

AI Support chats allow you to create dedicated chats which are published to volunteers, trained by you and allow volunteers to ask questions and receive instant responses. The aim is to improve the volunteers access to information without having to search or browse and save time for your staff team that is currently spent responding to volunteer queries.

Creating an AI chat

Go to the section on the left of Volunteero labelled "AI Support". Press "+ AI Chat".

Add the required information accordingly:

  • Chat Name and Chat Description - This is what volunteers will see. Consider what name of the chat will prompt volunteers to use it and as relevant questions for that chat.

  • Groups - You can if you would like, create specific AI support chats for specific groups of volunteers. If a volunteer is in given group, they will see any support chats available to them (the same if they are in multiple groups). The same as missions if you don't select a group it will go to All Volunteers.

Pro Tip 🧠 For most organisations, one chat will suffice. The underlying technology is smart enough to understand the volunteer query, whether that be from someone doing one role or another, and provide a great response. It is all about the context of "training" behind it.

Training your AI Chat

Training is essential, the AI chat doesn't know what it doesn't know. The quality of the responses it provides volunteers with will depend on the quantity and quality of the information it is trained on.

Please note, all AI chats are trained on standard Volunteero usage information by default so for the most part, you do not need to add context about how to use Volunteero. The only instances where you may wish to do so is where perhaps you ask volunteers to engage with the app in a very specific way for your organisation.

Alright, let's get into it.

Instructions

This is where you input general information for the chat that isn't related to specific volunteer queries. We provide some useful examples. These are not required but may improve the quality of the chat and ensure it behaves and responds in a fashion you're happy with.

See image below for the standard questions that we provide that will help your AI Agent.

Knowledge Base

This is where you provide resources that the chat is trained on and can draw from when responding to volunteer queries. There are 3 types of resources you can include:

  1. Knowledge Hub - These are your native Volunteero resources that you have created within the Volunteero hub, just press + next to the one(s) you wish to include for that chat. For more information about the knowledge hub, read our dedicated article here.

  2. Upload Document - This allows you to upload various types of documents like PDF and word documents.

  3. Snippet - This is your secret sauce when it comes to really fine tuning your AI chat. You can create a snippet where you can add information directly to a snippet which the AI chat draws from. Just press + snippet and add additional information to that snippet.

Note you can edit existing snippets. This is a great way to continuously improve on responses. Use the feedback (covered in more detail later in the article) to see what is working well and where there are gaps. If you notice similar questions coming up with poor or incomplete responses, edit your snippets to give the chat the right information for next time. Note it may take up to 2 minutes or so to take effect.

  • Available - You can see the availability status of your resources. If there is a tick, that means the AI has the latest information. If it shows a refresh icon, you can give it some time, refresh and you should then see a tick.

Pro tip 🧠 - Add a snippet called Volunteer FAQs and use this continuously to make improvements. We like the format of
Question:

Answer:

This just makes it crystal clear but it is not essential.

Preview (i.e. internal testing)

Before you publish your AI chat, we would recommend using the preview section. You can simply ask questions as if you are a volunteer.

Try asking in different ways, with different terminology. You are stress testing the chat to ensure that no matter how volunteers engage, the chat has the resources to provide a good response. Keep testing and adding snippets, articles and documents until you are happy with the answers your AI agent is providing.

Publish & Volunteer's view

When you've followed all the above steps and you're happy with the answers, you can press "Publish" and the chat will be live.

If your volunteer cannot see the chat it will be because either they haven't got the latest app version OR they are not in the group that corresponds to the group for that chat. This is what the volunteer will see.

The chat will be clearly marked as an AI chat and your description will be visible when they click on the chat.

Feedback

When a volunteer receives a response from the AI chat, they are prompted to give feedback in the form of πŸ‘ or πŸ‘Ž. They also have the ability to give optional text based feedback.

It you press on "Feedback" at the top of your AI chat, you will be able to see and filter all responses. You can press on a specific response on the left, you can view any feedback at the top and you can scroll down to see all questions and response.

We highly recommend periodically checking this and using the information to further optimise your chat with training through the knowledge base.

AI Chat FAQs

Question: What AI Model do you use for this feature?

Answer: This feature uses the AI Model, Claude.

Question: Is it damaging to the environment or what is the impact?

Answer: We are monitoring this closely but initial estimates show each response will use the same amount of power if not less than sending an email. In short you would see no comparable environmental impact compared to your current approach, if not a decrease.

Video Tutorial

As always, if you require any further assistance with this function, please use the support button in your Staff Dashboard or email us at support@volunteero.org. Our team will be happy to help! 😊

Did this answer your question?