Build, Train and Test Conversation AI

From the main Ivorey menu, go to Settings > Conversation AI > click the [Create Bot +] button, in the top right corner > then select to create either a [General Q&A] or [Appointment Setting] bot

Note: For the purpose of this article, we'll focus on the General Q&A bot (most common). ​

In the Brand Voice section, insert your brand name, business information (e.g. business hours, website, etc.), and set the brand tone of voice according to your desire

1. Under the [Additional Options] tab, we strongly suggest adding further information about your target audience
   ​
   Note: Keep the language as simple, clear and as easy to understand as possible.
   
   You can feed your answers to ChatGPT and ask it to re-write it, using this bonus prompt: "I have written information about my target audience below. I will be using this data to train my AI agent on my brand. Please rewrite this content so it's clear, direct, and easy for an AI agent to understand and use in conversations".
   ​
   If you have multiple target audiences or language styles, you can set up multiple brand voices. This is helpful if, for example, you want to use one bot to provide support to your clients, and then another to respond to leads on social media. ​

In the Bot Settings section:​

1. Give your AI Bot a Name - For example, your first bot could be for responding to social media DM's, so you might name it "Social Media Bot"​

2. Select the Channels you want the bot available on - The options are Instagram, Facebook, SMS, Chat Widget SMS, Live Chat or WhatsApp (if you have a paid WhatsApp subscription)​

3. Select [Make Primary Bot], if it should be used as the main bot that replies to messages on that channel, and set the Wait Time before the bot should respond

4. Enter the First Message that a bot should send to users - In the age of AI, we suggest transparent statements that make it clear the user is speaking to a bot, although this is entirely up to you​

5. Insert the Goals of the bot - For example, in a social media conversation the main goal could be to generate leads. Therefore, you would prompt the bot to collect the name and email address. Once collected, it will populate those fields against the contacts record in the Ivorey CRM​

6. Tick the [Skip if already filled] checkbox, if you don't want the bot to ask the lead for information the system already has stored​

7. Toggle ON [Appointment Bookings], if you want the bot to be able to directly book appointments

   1. Select the exact calendar that appointments should be made in (e.g. Discovery/Sales call calendar)

   2. Select the workflow that the lead should be enrolled in after booking (this will need to already be set up). The workflow can include steps like tagging the contact and sending a notification/confirmation email to you and/or the contact
      
      Note: If someone already has an active appointment, the bot will stop trying to book and switch to Q&A mode.​

8. Select [Receive Email Notification if The Bot Doesn't Know the Answer] > and choose which team member you'd like to receive these notices​

Select the Mode of the new AI bot you just built: ​

1. Off – Bot is inactive. You can still train and test it, but it won’t respond to contacts (this is the suggesting starting phase)​

2. Suggestive – Bot suggests responses you can edit or send manually (after rigorous testing, we strongly recommend moving to this stage before setting the bot onto auto-pilot)​

3. Auto-Pilot – The bot replies automatically on your behalf (we recommend this as a final stage once all testing and quality assurance checks have been passed)

Now your bot is ready for testing

From the main Ivorey menu, go to Settings > Conversation AI > Bot Training > and add in your Sources:

1. Add public URLs (e.g. your website, help docs, etc.)

2. Create a Knowledge Base

3. Manually add Question & Answer pairs

From the main Ivorey menu, go to Settings > Conversation AI > Bot Trial

Choose the intent, either [Q&A] or [Booking]

1. Send test messages to see how the bot responds

2. Use thumbs up or thumbs down to give feedback and improve responses

Update and improve the Sources (e.g. knowledge base) to continue improving the responses