Build Your First No-Code AI Agent | Full Relevance AI Tutorial

Setting Up No Code AI Agents with Relevance AI

In this section, Ben introduces the topic of setting up no code AI agents with Relevance AI and shares his experience in building these agents for clients over the past six months.

Introduction to AI Agents and Relevance AI

• Ben discusses his journey of delivering around 15 AI agents to clients within a year, emphasizing that he is still learning but has gained significant experience in relevance AI.

• The video is structured into three parts: defining AI agents, exploring key concepts in Relevance Ai, and building an inbound lead manager agent. Viewers can skip to relevant sections using timestamps provided.

Understanding AI Agents

• Definition of an AI agent as software automating processes, making decisions, and interacting intelligently to achieve goals. Components include prompts, knowledge base access, tools for interactions with software.

• Differentiating between AI automation setups and AI agent setups based on decision-making capabilities. While both can yield similar outcomes, agents have more autonomy in tool selection for goal achievement.

Use Cases of AI Agents

• Highlighting use cases where human involvement is preferred such as customer service, sales reps wanting control over crafted emails, marketing content review needs high precision.

• Categorizing agents into co-pilot (human-in-the-loop supervision like content repurposing) and autopilot (fully automated tasks like prospecting and email responses), depending on client requirements.

Opportunities with AI Agents

This section delves into the vast opportunities presented by utilizing AI agents in businesses to enhance efficiency and differentiate from competitors.

Business Advantages of Using AI Agents

Automating Workflows with AI

In this section, the speaker discusses the significance of automating workflows using AI, highlighting its impact on businesses and individuals.

Importance of AI in Business

• Elon Musk emphasizes that AI-adapted businesses will excel in the near future.

• Non-tech individuals can now create powerful business applications using no-code solutions.

Relevance AI Features

• Relevance AI offers a no-code AI agent builder, distinguishing it from other tools like Crew AI.

• It provides templates and software integrations for CRM, LinkedIn scrapers, and website scrapers.

Agent Management and Tracking

• The platform offers a user-friendly UI for agent management within companies or client delivery.

• Users can track usage through dashboards and work with multi-agent systems for complex tasks.

Key Concepts in Relevance AI

This segment delves into essential concepts such as tools, agents, and multi-agent systems within Relevance AI.

Understanding Key Concepts

• Tools are rule-based sequences akin to make.com for executing tasks efficiently.

• Agents leverage tools to achieve complex goals while maintaining customization options.

Workflow Complexity

• Multi-agent systems allow intricate task execution by coordinating multiple agents effectively.

• Tools in Relevance AI operate based on rule-based sequences with access to variables, software APIs, and knowledge bases.

AI Tool Functionality

This part explores how AI tools function within Relevance AI through an example of building an email personalization tool.

Building Personalization Tools

• Users input data like names into variables to personalize emails effectively.

• Sequential steps involve Google search API calls, L&M analysis for LinkedIn URLs, and personalized email generation based on scraped data.

Tool Complexity Management

• Keeping tool complexity limited to around 10 steps ensures efficiency and reduces error rates.

AI Agent Features

The discussion shifts towards the features of AI agents in Relevance AI including triggering mechanisms from external software.

Enhanced Agent Capabilities

Core Instructions and Flow Builder

In this section, the speaker discusses the core instructions and the flow builder within the system, highlighting their significance in enhancing agent performance.

Core Instructions

• The core instructions serve as the system prompt for agents.

• The Flow Builder allows additional instructions to be provided for specific scenarios, improving agent performance.

Agent Labeling and Sub Agents

This part focuses on agent labeling and sub-agents, essential components for managing tasks efficiently within a multi-agent system.

Agent Labeling

• Agents can label tasks they have completed, such as researching leads or sending emails.

• Labels provide a clear overview of an agent's progress in handling tasks.

Sub Agents

• Sub-agents enable collaboration within a multi-agent system.

• Various settings related to sub-agents will be explored later in the video.

Client Case: B2B Financing Solution

The speaker introduces a client case involving a B2B financing solution from Denmark, emphasizing challenges faced with managing inbound leads and the impact on sales efficiency.

Client Background

• The client is a B2B financing startup from Denmark facing issues with managing inbound leads.

• Sales representatives were struggling to prioritize between SMB and Enterprise leads efficiently.

Pain Points

• Inefficient handling of inbound leads led to time wastage and low conversion rates.

• Sales reps were overwhelmed with bad leads, lacking preparation for calls.

AI Agent Solution

• An AI agent was developed to automate lead research, update CRM data, prioritize leads based on scores, identify decision-makers, and generate personalized call scripts.

AI Agent Tools Setup

Details about setting up the AI agent tools are discussed here, outlining how various tools work together seamlessly to streamline lead management processes effectively.

Tool Setup Process

• Eight different tools are integrated into the AI agent setup.

• Tools include lead researcher for information gathering, lead scoring for prioritization, CRM updates, decision-maker identification research, and personalized call script generation.

Relevance AI Dashboard Demo

A demonstration of how the Relevance AI dashboard operates is provided along with insights into setting up an AI agent using available templates for customization.

Dashboard Operation

• The Relevance AI dashboard triggers actions when new contacts are formed in HubSpot.

AI Inbound Lead Agent Process Overview

In this section, the AI inbound lead agent process is detailed, showcasing how it conducts lead research, scoring, CRM updates, and personalized call script generation.

AI Inbound Lead Agent Process

• The AI inbound lead agent initiates by utilizing the lead researcher tool to research and score leads. It performs tasks such as updating CRM with lead information.

• Progressing further, the agent tags conversations with leads and endeavors to identify decision-makers within organizations. A summary of all actions taken is provided at the end of the process.

• After finding the decision-maker, the agent proceeds to generate a personalized call script autonomously. This automated process streamlines lead management effectively.

Lead Enrichment and Outreaching Process

This segment delves into the outcomes of the AI inbound lead agent's activities, including creating new contacts, assigning lead scores for prioritization by sales reps, and generating personalized call scripts.

Lead Enrichment Details

• The system creates a new contact entry with details like lead score, LinkedIn account information, job title of the lead, company website URL, industry insights, decision-maker profiles including LinkedIn accounts and summaries.

• Additionally, a customized cold call script is formulated for each contact to enhance engagement. The script includes tailored greetings and spin selling questions aimed at initiating meaningful conversations with prospects.

Relevance AI Dashboard Features

Exploring Relevance AI dashboard functionalities encompassing pricing plans ranging from free accounts to paid options offering various credits per day for users' convenience.

Relevance AI Dashboard Insights

• Users can access different plans on Relevance AI starting from a free account providing 100 daily credits up to paid options priced at $200 with additional credit purchase capabilities at $2 per thousand credits. Beginners can leverage free credits effectively for initial AI agent development without significant cost concerns.

• Within Relevance AI's interface lies a repository of agents created by users featuring tools for knowledge uploads, templates availability including both general tools and specific agent templates catering to diverse needs efficiently. Analytics tracking and integrations with platforms like Google Hotspot and Slack are seamlessly integrated into the dashboard for enhanced user experience and functionality customization options through API key configurations ensuring cost optimization strategies are in place while using external services like OpenAI instead of relying solely on Relevance AI's pricing structure.

Managerial Tasks and Tools Overview

The section discusses the tasks a manager needs to perform efficiently, such as researching incoming leads, finding decision makers, updating CRM systems, and generating personalized call scripts based on research results.

Managerial Responsibilities

• Managers must follow specific rules when new contacts are formed in HubSpot, including starting and ending with designated tools.

• Emphasize the importance of researching decision makers using appropriate tools, especially when they differ from the original lead.

Tool Descriptions and Notes

• Providing detailed descriptions of tools used is crucial for agent performance; adding notes for specific issues enhances problem-solving efficiency.

• Instructions given at the beginning or end of prompts are more effective; incorporating notes can lead to quicker issue resolution.

Enhancing Agent Performance with Flow Builder

This part highlights the significance of utilizing the Flow Builder tool to ensure agents follow a structured workflow effectively.

Utilizing Flow Builder

• The Flow Builder allows for setting instructions in a sequential manner to guide agents through tasks efficiently.

• Incorporating conditions based on outcomes like finding or not finding decision makers helps tailor agent actions accordingly.

Optimizing Output with Decision Maker Handling

Discusses how handling decision maker information impacts agent workflows and call script generation.

Decision Maker Management

• Distinguishing between scenarios where decision makers are found or not found influences subsequent actions like updating CRM info and generating call scripts.

• Properly managing decision maker data significantly improves overall output quality by streamlining processes.

Labeling Tasks for Clarity

Explains how labeling tasks within the UI aids in tracking progress and ensuring completion of essential steps.

Task Labeling Benefits

• Assigning labels to completed tasks provides clarity on progress within complex workflows, enhancing task management efficiency.

Tools and Agents Management

In this section, the speaker discusses the management of tools within agents, highlighting the importance of balancing the number of tools to avoid confusion and inefficiency.

Managing Tools in Agents

• The speaker emphasizes that having around 10 tools for AI agents is optimal to prevent confusion and ensure effective decision-making.

• Introducing sub-agents as a solution for more complex processes, allowing for the creation of intricate workflows by adding specialized agents.

• Demonstrating how to deploy agents by sharing or embedding them, providing options for public access or integration into websites.

Building Tools for Agents

This part focuses on creating tools within agents, showcasing specific examples and explaining the process step by step.

Creating Tools

• Introduction to lead researcher tool as an example, emphasizing the importance of naming and describing tools clearly for agent understanding.

• Exploring the option to add a knowledge base to tools for enhanced functionality, with potential applications like personalized email generation.

• Defining user inputs in tools, highlighting their significance in providing necessary data points for tool operation.

Task Sequencing in Tools

The discussion delves into task sequencing within tools, outlining a rule-based approach to streamline processes effectively.

Task Sequencing

Extracting Company and Lead Information

In this section, the speaker explains the process of extracting company and lead information using a tool that prompts for email extraction, website identification, LinkedIn search, and data point extraction.

Extracting Company Website

• Prompt to extract website from email.

• Variable setup with double brackets.

• Domain name extraction with HTTPS addition.

Google Search and LinkedIn Extraction

This part focuses on utilizing Google search and LinkedIn scraping to gather relevant company information efficiently.

Output Generation

• Tool output showcasing the company website.

• Step-by-step demonstration of the tool's functionality.

• Storing company website in a variable for future use.

LinkedIn Research and Data Extraction

The discussion shifts towards leveraging LinkedIn for comprehensive company research and data extraction.

Utilizing LinkedIn Data

• Importance of finding company LinkedIn for research.

• Integration of Google search API to find LinkedIn profiles.

• Storing Google search results in a variable for further processing.

AI Instruction for LinkedIn URL Extraction

Instructions are given to AI tools to extract relevant data points from LinkedIn profiles efficiently.

AI Guidance

• Tasking AI with finding the most likely company LinkedIn URL.

• Detailed instructions provided within the tool interface.

• Emphasis on expert analysis of Google search results for accurate extractions.

Data Point Separation Using Json Format

The speaker elaborates on separating multiple data points efficiently using Json format within relevance AI tools.

Data Organization

• Need for segregating various data outputs effectively.

• Instructing L&M to output findings in Json format.

Manual Output Definition

In this section, the speaker explains the importance of manually defining the output variable to incorporate information from previous steps into the final output.

Manual Output Definition Process

• Defining the output variable is crucial as it requires all information gathered from previous steps for the final output.

Tool Functionality Overview

The speaker provides an overview of how tools work, demonstrating manual selection and showcasing options like 'last step' for obtaining only the last output.

Tool Functionality Demonstration

• Manual selection allows incorporating data from various steps into the final output.

• 'Last step' option enables extracting only the output of the last step if needed.

Advanced Settings and Tool Capabilities

Exploring advanced settings and capabilities of tools, including defining models, providers, and powerful functionalities like connecting tools without agents.

Advanced Tool Settings

• Tools offer advanced settings such as defining AI models like GPT-3 and selecting providers like Entropic or Google Grock.

• Tools can perform powerful tasks independently without agents by connecting or triggering actions between software applications.

CRM Integration with Lead Info Update

Demonstrating how to update CRM with lead information using APIs and integrating tools to enhance CRM data with lead details.

CRM Data Update Process

• Integrating tools with APIs to update CRM with lead information efficiently.

• Utilizing Hopspot API for updating contact details based on gathered research data.

API Call Setup for CRM Update

Detailing the process of setting up API calls for updating CRM contacts with specific lead information using variables and property definitions.

API Call Configuration

• Configuring API calls involves specifying methods like patch for updating contacts in Hopspot CRM.

• Defining properties within Hopspot CRM through internal names obtained from property settings.

Decision Maker Tool Setup

In this section, the speaker explains the setup of a decision maker tool using a company's URL and Google search to find key personnel in German companies.

Decision Maker Tool Process

• The user inputs the company URL obtained through lead researcher tool.

• Conducts a Google search for the company URL plus "impressum" as required by German law.

• The impressum page lists decision makers and CEOs legally mandated in Germany.

Finding Decision Makers in Germany

This part focuses on leveraging legal requirements in Germany to identify decision makers through website content and Google searches.

Identifying Decision Makers

• Utilize Google search with specific keywords to locate relevant website pages.

• Extract information using AI tools specialized in analyzing search results.

• Look for "Geschäftsführer," the German term for decision maker, on the website.

Extracting Key Information

The discussion centers around extracting crucial data from websites to identify decision makers efficiently.

Website Content Extraction

• Use web scraper tools to extract website content.

• Search for specific names like "Geschäftsführer" indicating key personnel.

• Analyze extracted data to pinpoint the most likely decision maker of the company.

LinkedIn Scraping and Data Analysis

This segment delves into scraping LinkedIn profiles of identified decision makers for further analysis and data extraction.

LinkedIn Data Extraction

• Run research tools on decision makers' LinkedIn profiles.

• Scrape LinkedIn data, including summaries, from identified profiles for comprehensive analysis.

Utilizing Knowledge Bases

Exploring how knowledge bases enhance AI capabilities by providing historical email replies and general company information for responses.

Knowledge Base Integration

• Incorporate FAQ about the company and past email replies as knowledge bases.

• Train AI to mimic writing style based on past email responses stored in knowledge base CSV files.

Setting Conditions for Tool Execution

Discusses setting up conditions within tools to ensure proper execution based on predefined criteria such as presence of specific URLs or content types.

Conditional Tool Execution

• Establish conditions determining when specific steps should run within the tool sequence.

Agent Setup and Integration with API

In this section, the speaker discusses setting up an agent with triggers using an API for integration.

Agent Setup and Triggers

• The API is used to connect to various software applications, allowing for flexibility in integration.

• Triggers are set up so that when a contact form is submitted, the contact information is sent to HubSpot.

• Integrating AI agents involves triggering them when new contacts are created in HubSpot.

Connecting CRM with AI Agent

This part focuses on connecting different CRMs like Pipedrive with the AI agent through HTTP requests.

Setting Up Connection with CRM

• When a new lead is added or updated in a CRM like Pipedrive, it triggers the connection to the AI agent.

• Configuring HTTP requests for relevance AI involves setting the method as post to trigger the agents.

Configuring Headers and Request Body

Here, details about configuring headers and request body for API connections are discussed.

Header Configuration

• Two headers need to be added: Content Type (application Json) and Authorization (API key).

• The request body should be set as raw content type JSON containing relevant data for the AI agent.

Customizing Request Content

Customizing request content by including variables from HubSpot for effective communication with the AI agent.

Customized Message Content

• Including variables such as HubSpot ID, email, and lead name in the message sent to the AI agent enhances data accuracy.

Finalizing Setup and Considerations

Final steps involve ensuring proper setup of API connections and understanding synchronous triggers' limitations.

Completing Configuration

• Confirming settings by selecting "part response yes" ensures successful setup of communication between systems.

Additional Considerations and Conclusion

Addressing asynchronous triggers' necessity for long-running tools and concluding remarks on video content creation.

Asynchronous Triggers

• Synchronous triggers have a time limit of 30 seconds; longer processes require asynchronous endpoints for continued operation.

Conclusion

• The video creator invites viewers to engage further with AI agent use cases while hinting at future tutorials on asynchronous endpoint setups.