AI: SMS-iT CRM AI Features

Navigate to https://aicpanel.smsit.ai/home

Click on "Artificial Intelligence"

1. Conversations Completed Total Conversations: Displays the number of completed conversations across all channels. Breakdown by Channel: Shows the distribution of conversations among text, social media, email, and chat. 2. Response Time Total Response Time: Indicates the cumulative time spent responding to conversations. Breakdown by Channel: Details the response time for each channel, such as text, social media, email, and chat. 3. Engagement Rate Engagement Rate: Measures the level of user interactions with the service, indicating user activity and interest. 4. Conversion Rate Conversion Rate: Tracks the percentage of conversations that lead to a successful outcome, such as a sale or sign-up. 5. Click-Through Rate (CTR) CTR: Shows the percentage of users who clicked on a link or call-to-action within a conversation.

6. User Statistics User Statistics: Provides detailed metrics about user activities and behaviors, such as user satisfaction, retention, and acquisition levels. 7. Trained Data Trained Data: Shows the volume of data used to train the AI models, indicating the system's learning scope and capabilities. 8. Error Rate Error Rate: Represents the number of errors or issues encountered in the system, important for monitoring system reliability. 9. Churn Rate Churn Rate: Measures the rate at which users stop using the service, indicating customer retention levels. 10. Resolution Rate Resolution Rate: Indicates the percentage of conversations that were successfully resolved, reflecting the effectiveness of support.

11. User Demographics Demographics: Provides a breakdown of users by country, helping to understand the geographic distribution of your user base. 12. Social Media Integration Social Platforms: Displays the number of users interacting through various social media platforms, such as Facebook, YouTube, LinkedIn, Twitter, Dribbble, and GitHub. 13. Performance Metrics - Average Session Duration: Measures the average length of user sessions, indicating engagement. - Language Accuracy: Assesses the accuracy of language processing in interactions. - Task Completion Rate: Shows the percentage of tasks successfully completed within conversations

13. Performance Metrics (Continued) - Intent Recognition Accuracy: Evaluates how accurately the system understands user intents. - Entity Recognition Accuracy: Measures the accuracy of identifying specific entities within conversations. - Sentiment Accuracy: Tracks the accuracy of sentiment analysis, indicating how well the system understands user emotions. - Average Temperature: Can be a metaphorical measure indicating system performance or user engagement level. - NLP Model Confidence Scores: Reflects the confidence level of the AI in its responses. - Naturalness of Responses: Assesses how natural and human-like the AI's responses are. 14. Visitors Visitors: Shows the number of visitors interacting with the system. 15. Session Dropout Rate Session Dropout Rate: Tracks the rate at which users drop out of sessions, indicating potential issues with engagement or satisfaction. 16. Traffic Sources Traffic Sources: Displays the origins of traffic to the system, such as direct, referral, or organic search, to understand how users are finding the service.

Click on "Settings"

"Choose your AI model, text" Select your preferred settings: - AI model - Text-to-voice type (ElevenLabs recommended) - Text-to-voice language - Voice model

Attach phone numbers to each AI model. Different models can have separate numbers assigned to them for incoming texts.

"Temperature: This controls the randomness" - AI Model Parameters: - Temperature: Controls output randomness (lower values increase focus) - Presence Penalty: Discourages repetition - Frequency Penalty: Encourages use of less common words/phrases - Top P (nucleus sampling): Sets probability threshold for token selection

Additional options:

"Include Trained Data" checkbox "Train" button for incorporating custom data

These parameters allow fine-tuning of the AI model's output characteristics.

Click Save

Adding an API Key for ChatGPT: - Click on the ChatGPT icon to open the API key configuration section. - This will allow you to enter your ChatGPT API key for integration with the SMS-iT CRM system.

Note: This example uses ChatGPT, but the process may be similar for other AI models or services supported by our platform.

Adding Your API Key: - In the API key configuration section, you'll see a field to enter your API key. - Carefully type or paste your ChatGPT API key into the provided field. - Click the "Submit" button. - This action will save your API key and integrate ChatGPT with your SMS-iT CRM system.

Remember: Keep your API key confidential and secure. Never share it with unauthorized individuals.

Click on "Custom AI Persona"

Creating Your AI Persona When setting up your AI Persona: - Choose a descriptive name for easy reference - Create multiple personas for different aspects of your business (e.g., sales, marketing, customer service) - Each persona can be specialized to handle specific tasks or departments

SMS-iT offers flexibility in AI model selection: - Choose from ChatGPT versions 3.5, 4, or 4o - Claude AI is also available as an option (coming soon) - SMS-iT uses a BYOK (Bring Your Own Key) model - Enter your GPT API key for access This approach provides more options and can help reduce costs

Training Your AI Persona To customize your AI Persona: - Add your website content - Include frequently asked questions and answers - Input any relevant business information - You can edit and add to this information at any time

Testing Your AI Persona:

1. Simulate customer interactions:

• Compose text messages as if you were a potential customer or client.
• Ask a variety of questions that your target audience might have.

2. Evaluate response accuracy:

• Review the AI's answers for correctness and relevance.
• Check if the information provided aligns with your business offerings and policies.

3. Assess tone and style:

• Ensure the AI's communication style matches your brand voice.
• Verify that responses are appropriate for your target audience.

4. Make necessary adjustments:

• Identify areas where the AI's responses need improvement.
• Refine the AI's knowledge base or adjust its parameters as needed.
• Retrain the model if significant changes are required.

5. Repeat the process:

• Continue testing with different scenarios and question types.
• Regularly update and fine-tune your AI persona to maintain accuracy and relevance.

By following these steps, you can ensure that your AI persona accurately represents your business and effectively assists potential customers or clients.

Click on "AI Transfer"

Setting up AI Transfer: - Add a keyword to trigger the AI transfer - Include the question from the AI chat that will trigger a response - Add one or more numbers that will receive the question - Provide a brief overview of the customer

Configuring AI Transfer Message:

Compose the message to be sent from AI to human Use merge tags for personalization: - First name - Mobile number - Brief outline of the AI conversation

This allows the human to respond via text or call with context

Save or Update AI Transfer:

Click "Save" when creating AI Transfer for the first time Click "Update" when editing existing AI Transfer settings

Click on "Train"

AI Training Data Sources: - Website content - Text documents - Chat logs - User feedback - Industry-specific information - YouTube videos - PDFs - Sitemaps - DOC files - Code documentation - Notion content

This variety of sources allows for comprehensive AI training tailored to your business needs.

Website Data Extraction:

Add exact URL for specific page extraction Add domain for extraction from all URLs within the domain

Click on "Extract Data"

Data Extraction Options:

"Select all pages" for comprehensive extraction Individually select from "Existing pages" for targeted extraction

Click on "Extract Data"

Your data is now imported

Select Data Analysis Method: Choose your desired mode of analysis from the "Choose Data Analysis" dropdown menu to train your AI assistant effectively.ta

Intent Classification:

Purpose: To categorize user inputs into predefined intents or actions. Process:

Analyzes text data to identify patterns and key phrases associated with specific intents. Groups similar queries or commands into distinct categories.

Training focus:

Teaching the model to recognize various ways users might express the same intent. Building a robust set of intents that cover the range of user requests.

Use case:

Ideal for creating chatbots or AI assistants that need to understand and respond to specific user requests. Helps in automating responses to common queries or actions.

Contextual Awareness:

Purpose: To extract and utilize contextual information from conversations or user interactions. Process:

Analyzes sequences of messages or user interactions to understand context over time. Identifies relationships between different pieces of information within a conversation.

Training focus:

Teaching the model to maintain and use conversation history. Developing the ability to understand references and implicit information.

Use case:

Useful for creating more sophisticated AI systems that can handle complex, multi-turn conversations. Enables personalized responses based on user history and conversation context.

Emotion Analysis:

Purpose: To detect and categorize emotional content in text data. Process:

Analyzes text for emotional indicators, sentiment, and intensity. Categorizes text into emotional states or sentiment categories.

Training focus:

Teaching the model to recognize emotional language and tonal indicators in text. Developing sensitivity to nuances in emotional expression.

Use case:

Valuable for sentiment analysis in customer feedback or social media monitoring. Can be used to prioritize or route messages based on emotional content.

When you've decided on your data anaylsis model Click on "Train"

Your AI assistant has been trained

Enhancing Your AI Assistant's Knowledge: Continuously improve your AI assistant by utilizing various data import options to: - Expand its knowledge base - Refine understanding of specific topics - Keep it updated with the latest information

Regular incorporation of new data from diverse sources ensures your AI assistant remains: - Current - Relevant - Increasingly capable of addressing a wide range of user queries and tasks

Here's a brief explanation for each data import option:

1. Website: Extracts and processes content from specified web pages or entire websites.

2. Text: Allows direct input or upload of plain text documents for training.

3. Chat Logs: Imports conversational data from previous chat interactions or customer service logs.

4. User Feedback: Incorporates customer reviews, comments, or survey responses into the training data.

5. Industry Specific: Focuses on importing specialized data relevant to a particular industry or field.

6. YouTube: Extracts information from YouTube video transcripts, titles, and descriptions.

7. PDF: Imports and processes text content from PDF documents.

8. Sitemap: Uses a website's sitemap to systematically import content from all listed pages.

9. DOC: Imports and processes text from Microsoft Word documents.

10. Code Doc: Specifically designed to import and analyze software documentation or code comments.

11. Notion: Imports content from Notion workspaces, including pages and databases.

These options provide a wide range of data sources to train your AI, allowing you to customize its knowledge base according to your specific needs and industry.

Click on "Personalize AI"

Click on "+ Add Q & A"

Insert a question

Click on "Add"

Click on the arrow to add a response

Click on "Edit"

Add your response

Click on "Save"

Customizing AI Responses: - Edit existing questions or responses - Add new questions and answers as needed - Review chats, emails, and other sources to identify frequently asked questions - Continuously update to improve AI performance - Add as many Q&As as necessary to thoroughly train your AI

This process ensures your AI assistant can effectively address common customer inquiries and provide accurate, relevant information.

Managing Training Data: Access all training data by clicking on the appropriate import channel Retrain data as needed to keep your AI up-to-date Regular review and updates ensure optimal AI performance

Retraining Process: To retrain any dataset: 1. Click on "Train" for the specific data you want to update 2. Review and modify the existing data as needed 3. Choose a different Data Analysis method if desired 4. Click "Train" again to implement changes

This process allows for continuous improvement and refinement of your AI assistant's knowledge and capabilities.

Retraining Existing Data: - Existing data appears in the text box - Add or remove information as needed - Option to use a different Data Analysis method - Click "Train" to retrain with updated information

This flexible approach allows you to keep your AI assistant's knowledge current and relevant to your business needs.