AI is not just a buzzword at RevCent. We are all-in on AI and have built features for our users to combine AI with their RevCent account. AI is becoming a core component of RevCent, bringing advanced capabilities and automation throughout many aspects of the RevCent ecosystem.
Communicate directly with the AI via text or voice within the RevCent web app, including the ability to generate charts using the AI Playground. Over 100 system tools are made available to the AI.
Run tasks autonomously in the background, communicating with the AI based on the design of your thread. Over 100 system tools are made available to the AI.
If you use chat GPT, you can link your RevCent account with chat GPT through the RevCent GPT. Perform a select number of System Tools specific to your RevCent account using the OpenAI GPT.
System tools are a core component within RevCent's AI. System Tools allow AI to communicate and take action within your RevCent account.
Many of the operations listed on the API V2 Docs are available to the AI as System Tools. AI is intelligent enough to know what endpoint to call, fields that are required, etc. The AI will then run the System Tools(s) on its own and provide your Web Chat or AI Assistant with the result. Over 100 system tools are available to the AI via Web Chat and AI Assistant. The RevCent GPT is limited to the number of tools due to OpenAI limitations on custom GPT's.
When chatting with the AI or the running of an AI Assistant, the AI is provided a list of System Tools it is permitted to use. The AI will determine any tool(s) it may need order to fulfill either an AI Chat, AI Assistant or GPT request.
Example 1:
Lets say you are chatting with the AI and you asked it to find a customer. The AI will read available System Tools, find the Search Customers operation and then make the Search Customers call using the RevCent API in order to conduct the search. The results will be returned to the AI and it will present them accordingly.
Example 2:
An AI Assistant step tells the AI to create a note and email the customer. The AI will read available System Tools, find the Create Note and Send SMTP Message operations. The AI will then make both API calls simultaneously or in order. The results will be returned to the AI and the AI Assistant will proceed.
When setting the trigger as API/AI for an Email Template or RevCent Function you will have the option to add one or more Custom Arguments. Custom Arguments allow the AI to pass dynamically generated information into either an Email Template or RevCent Function.
Before triggering an Email Template or RevCent Function, the AI will read available custom argument descriptions, generate a value for each argument and then set the name and value within the custom_arguments object. This allows you to have the AI use its own logic to generate content the email template or function will receive.
Each custom argument you add requires you to specify the name and description. It is important to understand both the name and description settings below.
The custom argument name is how you will access the generated content within an Email Template or RevCent Function. The name must be unique across all custom arguments. If you wanted to include the custom arguments in an email template, you would access it via the {{custom_arguments.*}}
handlebars in your template, where * is the name of a custom argument. If utilizing custom_arguments in RevCent functions, the custom arguments will be accessible via event.data.custom_arguments.*
within the function, where * is the name of a custom argument.
The custom argument description determines the content. The description of a custom argument is the most important part. The description is the “instructions” the AI will read in order to generate the content when triggering an Email Template or RevCent Function.
Important: Write the description of a custom argument in a way that the AI will understand what you want generated and outputted. The description should be written such that a human can understand exactly what you want generated. If the description is too vague, the AI may not generate the exact content you want.
Below are examples of custom arguments, including naming and describing.
Objective: You want the AI to determine what timezone a customer lives in.
custom_arguments.customer_timezone
Objective: You want the AI to generate a sale summary.
custom_arguments.sale_summary
Below is an example output for the custom arguments with the AI generated content/value for each argument
{
"custom_arguments": {
"customer_timezone": "US/Eastern",
"sale_summary": "The customer purchased a USB as well as a Laptop, for a total amount of $2,239.21"
}
}
Create, edit and delete AI Directives by visiting the AI Directives page at https://revcent.com/user/ai-directives
AI Directives give you the ability to automatically tell AI specific rules, caveats and business logic, thus helping the AI understand your business and data structure. This is extremely useful for chatting, instructions and report building. Instead of having to constantly explain internal logic to the AI, simply create AI Directives and they are automatically provided to the AI upon initiation of a Web Chat or an AI Assistant.
Each AI Directive you add requires you to specify the name, status, use and content. It is important to understand the settings below.
The custom argument name. This is for your own internal use and is not provided to the AI.
Enable or disable an AI Directive. Only enabled AI Directives are provided to the AI.
You have the ability to limit a directive to either Web Chat, AI Assistant or both. Useful when you want a directive specific to a method of use.
The content is what is provided to the AI upon initiation of a Web Chat or AI Assistant. The AI will read the directive first before initiating a conversation. Provide the AI with the rules, instructions or business logic you want it to comprehend. Write your content such that a human can also understand.
Important: The content should explicitly state when to follow and/or implement the directive. AI can sometimes implement a directive when it is unnecessary, leading to data misinterpretations. Check the examples below on how to create the content of a directive, including telling the AI when it should implement the directive.
Below are example AI Directives showing how the content should be written based on business logic and/or data structure.
Objective: You want the AI to know how to read and/or query data to determine upsells versus non-upsells. Instead of constantly having to tell AI how to determine an upsell, this explainer will resolve the issue. You will then be able to ask about upsells, such as “Give me the total sale gross for upsells today.” Notice how we tell AI to only implement this directive when we ask or refer to upsells.
An upsell is distinguished via metadata where name = is_upsell and value = true
A non-upsell, or initial sale is distinguished via metadata where name = is_upsell and value = false
Only distinguish between upsell vs non-upsell when explicitly asked about upsells.
Objective: You want the AI to know how to read and/or query data to determine affiliates. Instead of constantly having to tell AI how to get an affiliate ID when chatting or generating reports, you can instead simply say “Group by total sales gross by affiliate”. Notice how we tell AI to only implement this directive when we ask or refer to affiliates.
An affiliate or partner is the metadata value where metadata where name = affiliate_id
Only implement when explicitly asked about affiliates and/or partners.
Objective: You want to be able to ask about campaigns by name, in which the AI will need a campaign ID generate reports. The AI will attempt to filter by the name you give, however it should instead get the ID first based on the name you provided.
When asked to filter or query by campaign, first get the list of all campaigns, then find the ID of the relevant campaign(s).
Only do this when specifically asked about campaign(s).