The creation of artificial intelligence agents is an innovative and powerful feature offered by Tayon. The creation of the agent depends on several stages, and the goal of this article is to detail each one of them.
Requirements
To create an AI agent you will need:
– A Tayon account on the Pro plan or higher.
– An exclusive email account for your AI agent. It can be a Gmail account, for example. Ideally, it should be an email account with the domain of the company the AI agent represents. You can also ask your Tayon representative about our in-house email solution for AI Agents.
– If the AI agent is going to use WhatsApp, a WhatsApp number for the agent is also necessary.
– If the volume of messages the AI agent will send is very high (above 1000 messages per month), you will need an API key from OpenAI.
– Content about the company or offer to train the agent.
– A clear and specific goal for the AI Agent.
– Planning of the message sending flow and the response/return flow.
Creating the AI Agents Group
The first step is to create the AI agents group. The group can contain one or more agents.
In addition to a name for the group (free-form), you must select the type “Artificial Intelligence” for the group type. You can also set a maximum number of emails and instant messages that any agent in this group can send. In the instructions, you can create a specific prompt for the AI agents in this group. The prompt/instructions will be added to the prompts with the AI rules that we will discuss later.
Creating the AI Agent
The AI agent is nothing more than a member of your team, but one that is part of a group of the Artificial Intelligence type.
Setting the Language and Connecting the Email Account
The next step is to log into Tayon as the AI agent (use the email and password you just created to log in). After logging in, you must set the desired language (English, Portuguese, or Spanish). The agent will build the email and text messages based on the language of its profile. After setting the language, it’s time to connect the agent’s email account to Tayon.
Creating the Rules/Prompts for the AI
AI agents need specific information to create the content for emails, instant messages, etc., as well as instructions on how to respond to emails, messages, and how to behave toward your leads. If your company is a formal company, such as a financial institution or bank, for example, the AI agent needs to be formal. If your company is a marketing agency, the approach can probably be more relaxed.
The following rules can be created to guide your AI agent:
IMPORTANT: The names of the rules must be exactly as shown below; otherwise, the agent will not consider the rule when generating content for conversations. All rules are optional, but the lack of information may cause the AI agent to create messages that are not aligned with your sales objectives.
AI-About
This rule explains to the AI agent what your product or offer is. Avoid inserting too much information; provide only the basic and sufficient details so that the agent has a general idea of your business or product. Below is an example we use for Tayon:
Our AI platform offers these benefits:
*** Lead Capture ***
Capture leads without complications with our Exclusive Conversational Websites.
Eliminate forms and improve conversions while creating amazing experiences for your audience. With our conversational websites, you can capture leads at events such as exhibitions and personal networking meetings simply by sharing a QR Code.
Give your leads the option to take a picture of their business card and instantly see their contact information saved in our CRM. Use our native calendar system to schedule meetings and calls. This is an uncomplicated way for clients to schedule meetings with you.
Use our comprehensive CRM to manage your leads and work with your team.
You can also integrate Tayon.ai with almost any other system or platform.
*** AI Sales Agents ***
Prioritize contacts, generate and send personalized messages across various channels, and accelerate engagement with priority leads — 100% autonomously.
As soon as leads show interest, the AI Agent informs your sales representatives that they have a high-quality lead, ready to close.
AI Sales Agents work 24 hours a day. Without adding more personnel or outsourcing to an external company, you can start engaging leads immediately and schedule meetings 100% autonomously.
Revive stalled leads and transform your lead backlog into new conversations.
Ensure that every lead received gets timely and relevant follow-up.
Spend less time searching for leads and writing unsolicited emails and more time calling engaged leads who are ready to talk.
Agents autonomously handle active prospecting and follow-ups and manage all received leads simultaneously. They also ensure that the message remains consistent in all interactions with potential clients.
As you can see, we have highlighted only the main benefits of the platform. The AI agent itself will complement what is necessary to explain how Tayon works and why the lead needs our platform to improve their results.
AI-Instructions
The AI-Instructions rule is usually very short and to the point. It consists of instructions regarding the behavior and profile of the AI agent and, mainly, the objective you want the agent to achieve. Example:
You are Ayrton Ignácio, our sales representative. Your goal is to convince our lead to schedule an initial meeting with our sales manager for a presentation of the platform.
AI-Instructions-first
This rule is specific to creating the first contact message with the lead. Usually, the first message is the most important and needs to serve as an introduction. Here’s an example we use at Tayon:
We just met this lead at an event, and he/she (try to determine the lead’s gender by their name) provided us with their contact information. Please write a message saying it was great to have him/her interested in our AI solutions to enhance his/her business, and that we are looking forward to speaking with them soon about the possibilities our platform brings. Use some of the benefits in your message. Sign the message with your name and just below your name: Tayon AI Sales Agent.
AI-Instructions-followup
This is used for creating follow-up messages. Example:
This is a follow-up message, so mention the message you sent earlier and use different arguments to convince the lead to schedule a meeting. The link to schedule the meeting is https://contact.tayon.ai?ctt=$CRMDATA.UID
AI-Instructions-subject
Instructions for creating email subjects. Example:
Create the subject of the message based on the email content, but try to use a question to spark interest. It is also important to use the lead’s name in the subject if available.
AI-Instructions-response
Used to give specific instructions for responses to emails or messages received in the AI agent’s email inbox. For example, if one of the leads sends a question about the offered product or service. In this case, this instruction will only complement the knowledge provided in the AI-About and AI-Instructions rules. Example:
Reply to the message in a friendly manner and offer to schedule a Zoom session to clear all of the lead’s doubts.Creating the Task Flow for the AI Agent
For the AI agent to work, you need to create a task flow for it. To do this, create a new automation flow and select “AI Tasks” as the flow type/trigger.
1 – Select AI Tasks as the trigger/type of flow;
2 – To send emails, messages, or perform tasks that depend on the lead’s data, use the identifier $CRMDATA to retrieve information (see CRMDATA VARIABLES in Using Variables in the Flow);
3 – Use the identifiers for generating content with AI to create the subject. You can also set a fixed subject if desired, but it is recommended to use AI to generate the subjects and texts of the messages.
4 – Use the AI identifiers to create the body of the email.
5 and 6 – Create the next steps in your flow.
VARIABLES FOR CREATING CONTENT WITH AI
$EMAILDATA.SUBJECT
Creates the subject of the email according to the AI rule “AI-Instructions-subject.”
$EMAILDATA.AIFIRST
Generates the content of the initial contact email according to the AI rule “AI-Instructions-first.”
$EMAILDATA.AIFOLLOWUP
Generates the content of the follow-up email according to the AI rule “AI-Instructions-followup.”
$EMAILDATA.AIRESPONSE
Generates the content of the response email according to the AI rule “AI-Instructions-response.”
$IMDATA.AIFIRST
Generates the content of the initial instant message contact according to the AI rule “AI-Instructions-first.”
$IMDATA.AIFOLLOWUP
Generates the content of the follow-up instant message according to the AI rule “AI-Instructions-followup.”
$IMDATA.AIRESPONSE
Generates the content of the instant message response according to the AI rule “AI-Instructions-response.”
Creating the Flow for Handling Responses
When the AI agent sends an email or an instant message, the lead may reply, and that response needs to be handled according to your objectives. In some cases, the AI agent may send an email suggesting the scheduling of a meeting, for example, but not send the scheduling link. If the lead responds to the email saying they accept the meeting, you can send another email with the link, or perhaps you prefer that someone from the team gets in touch with the lead to schedule it.
To make it possible to handle these responses, all messages received by the AI Agent (whether by email or instant message) are analyzed by AI and classified into one of the following categories:
accepted_offer: The offer/suggestion made by the AI Agent was accepted.
quote: The lead requested a quote or proposal.
referral: The lead indicated another person responsible for the agent to contact.
call: The lead requested to receive a phone call from the agent/company.
sms: The lead requested that the agent send a message via SMS or WhatsApp.
complaint: The lead complained about the message or asked not to receive further messages (in this case, the AI agent will block sending new messages to that lead).
follow_up: The lead asked the agent to contact them again in the future.
automated: The response is an automated message.
more_info: The lead asked for more information to make a decision.
interested_not_ready: The lead indicated interest but said they are not ready to make a decision yet.
feedback: The lead sent some comment, criticism, or praise about the content of the message.
unsubscribe: The lead requested to be removed from the list and not receive any further messages (in this case, the AI agent will block sending new messages to that lead).
forwarded_internal: The lead forwarded the message to someone within the same company.
dissatisfied_current: The lead indicated they are dissatisfied with their current provider (possibly your competitor).
not_interested: The lead said they are not interested in your offer/product.
already_customer: The lead said they are already a customer of your company or a user of your product/service.
out_of_office: Message indicating that the lead is not available to reply to emails.
technical_issues: The lead indicated that they had some technical problem with your message (could not read it, error occurred, etc.).
email_nonexistent: Return message indicating that the email does not exist or that the lead no longer works at the company.
email_full: Return message indicating that the lead’s mailbox is full.
spam_challenge: Message indicating a challenge from some anti-spam system used by the lead.
bad_experience: The lead indicates they had a problem or a bad experience with your company/product/offer.
question: The lead asked something about your company/product/offer.
not_found: It was not possible to classify the message.
To create the Responses flow, simply create a new automation flow and select the trigger according to the type of message (Email, SMS, or WhatsApp):
The first step in this type of flow is generally a decision to identify the category of the response and execute the appropriate actions.
To indicate that you want to retrieve the value of the message category to compare it with one of the possible values of the list of categories, just select the values like in the image below. If you need to access the email category in any other part of the flow, you can use the variable $EMAILDATA.CATEGORY (for email response flows) or $IMDATA.CATEGORY (for SMS or WhatsApp response flows).
After the decision, your flow can continue with more decisions or actions. The flow below is an example of a response flow:
Besides the category variables, you can also use in your flow the variables:
$EMAILDATA.BODY
Contains the body of the received email message.
$EMAILDATA.EMAIL
Contains the email of the sender of the received email message.
$EMAILDATA.SUBJECT
Contains the subject of the received email message.
$EMAILDATA.CC
Contains email addresses that were copied on the received email message.
$IMDATA.BODY
Contains the text of the received instant message.
$IMDATA.PHONE
Contains the phone number of the sender of the received instant message.
You can also use all the $CRMDATA variables/identifiers to retrieve information about the lead (see CRMDATA VARIABLES in Using Variables in the Flow).
Putting the AI Agent to Work with Your Leads
For your AI agent to start working with your leads, simply associate leads to the agent in the CRM. You can associate leads one by one by editing the lead and selecting the AI Agent as the lead owner, select several leads in the leads list and associate them with the agent, or use the “Assign lead” action in the Automation and Conversational Websites flows.
IMPORTANT INFORMATION:
1 – When associating a lead to the agent, the agent will start sending messages according to the AI Task flow immediately, unless you have set a work schedule for the agent (recommended to avoid sending messages during non-business hours, for example) and it is outside the determined hours. You can configure the agent’s work schedule in the icon in the Team Members list.
2 – To avoid the agent’s messages being blocked by providers, the message sending cadence is reduced. The agent is capable of processing the sending of about 300 messages per hour, approximately. Therefore, when delivering a lead to the agent, it may take a few minutes before the first message is sent.
3 – If the email account you are using for the agent is from a mainstream email provider, the account will likely have daily/hourly sending limits. Check what these limits are and try to balance the agent’s message sending based on these limits. Remember that if the agent is programmed to send a first message and then send follow-up messages every X days, the volume of messages may accumulate and exceed the limits. For large quantities of leads, we recommend setting up an SMTP service separate from the email provider, such as PostMark, Sendgrid, Amazon SES, Mailgun, Mailjet, MailSender, etc. This solves the issue of sending limits imposed by providers.