Build a No-Code AI Agent to Automate Your Proposals | EASY Set-Up

Key Concepts

• No-code AI agent
• Meeting transcript automation
• Personalized proposal generation
• Invoice creation
• Slack notifications
• Relevance AI (no-code AI agent builder)
• Make.com (workflow automation tool)
• Fireflies.ai (meeting transcription service)
• GPT-4o (Language Model)
• Webhooks
• API Integrations
• JSON formatting
• Google Docs integration
• Stripe integration
• Agent prompting and SOPs (Standard Operating Procedures)
• Escalation to human approval

Agent Overview and Demo

The video demonstrates how to build a no-code AI agent that automates the process of generating personalized proposals and invoices from sales call transcripts, and also summarizes internal meetings. The agent uses Fireflies.ai to capture meeting transcripts, Relevance AI to process the transcripts and generate outputs, and Make.com to integrate with third-party services like Google Docs, Stripe, and Slack.

The demo showcases the agent processing a sales call transcript, extracting key information like client name, project details, budget, and proposed solutions. The agent then sends a Slack notification for approval before generating a proposal and invoice. After approval, the agent generates a personalized proposal using Google Docs, creates an invoice using Stripe, and sends an email to the client with links to both.

System Setup Overview

The system comprises a Relevance AI agent and Make.com workflows. The agent uses several tools:

1. Categorize Transcript: Classifies the meeting transcript as a sales call, internal meeting, or other.
2. Sales Info Extractor: Extracts key data points from sales call transcripts for proposal generation.
3. Human Approval Tool (Escalate to Humans): Sends a Slack message for manual review and approval before proceeding. This is a built-in tool in Relevance AI.
4. Proposal Generator: Creates a personalized proposal using Google Docs via Make.com integration.
5. Stripe Invoice Generator: Generates an invoice using Stripe via Make.com integration.
6. Summarize and Send Slack Message: Summarizes internal meeting transcripts and sends them to a Slack channel.

Step-by-Step Breakdown

1. Setting up the Make.com Trigger and Fireflies.ai Integration

• Webhook Creation: A webhook in Make.com is created to trigger the automation when a meeting transcript is available.
• Fireflies.ai Configuration: The webhook URL from Make.com is added to Fireflies.ai's developer settings. Fireflies.ai is configured to trigger the webhook upon transcription completion. The "Transcription completed" event is selected.
• API Key Retrieval: The Fireflies.ai API key is copied from the developer settings.
• Connecting Fireflies to Make.com: The API key is used to connect the Fireflies.ai account to the Make.com module.
• Meeting ID Variable: The meeting ID, sent by Fireflies.ai to the webhook, is used as a variable in the Make.com scenario.
• Iterator Module: The iterator module is used to separate the sentences array from the Fireflies transcript.
• Text Aggregator Module: The text aggregator module combines the speaker name and raw text from the transcript into a single text.

2. Relevance AI Agent Configuration

• Agent Profile and System Prompt: The agent is given a role, objective, and SOP (Standard Operating Procedure) using a detailed system prompt in markdown format. The prompt defines the agent's role, objectives, and the order in which to use the tools.
• SOP Importance: The SOP guides the agent on which tools to use based on the transcript categorization.
• Tool Descriptions: Each tool is described with its function, inputs, and outputs.
• Examples: Examples are provided to guide the agent's behavior in real-world scenarios.
• Notes: Notes are used to reinforce important rules, such as always using the human approval tool before generating a proposal.
• Flow Builder: The flow builder is used to visually define the decision tree for the agent, specifying different paths for sales calls and internal meetings. Conditions are added to determine the category of the transcript.
• Escalate to Humans: The built-in "Escalate to Humans" tool is used to send a Slack message for approval before generating the proposal and invoice.
• Model Selection: GPT-4o is used as the language model for the agent.

3. Tool Configuration

3.1. Categorize Transcript Tool

• Input: The entire meeting transcript.
• Output: The transcript is categorized as either a "sales call" or "internal call".
• AI Step: An AI step with a prompt instructs the language model to categorize the transcript.

3.2. Sales Information Extractor Tool

• Input: The entire sales meeting transcript.
• Objective: Extract relevant data for proposal and invoice generation.
• AI Step: An AI step with a prompt instructs the language model to extract specific information like client name, email, project details, and budget.
• JSON Output: The extracted information is formatted as a JSON object.
• Convert to JSON Module: A "Convert to JSON" module converts the text-based JSON output from the language model into a structured JSON format.

3.3. Google Docs Proposal Generator Tool

• Input Fields: Client name, solution proposal, project name, and budget.
• API Call to Make.com: An API call is made to a Make.com scenario to generate the proposal.
• Webhook URL: The same webhook URL used for the Fireflies.ai integration is used for this API call.
• JSON Payload: The data is sent to Make.com as a JSON payload.

3.4. Stripe Invoice Generator Tool

• Input Fields: Client name, client email, and amount.
• URL Parameters: The data is sent to a Make.com scenario via URL parameters.

3.5. Send Email Tool

• Input Fields: Proposal link, client first name, email, and Stripe invoice link.
• Gmail Integration: The tool uses the Gmail integration to send an email with the proposal and invoice links.
• Email Template: A standardized email template is used.

3.6. Summarize and Send Internal Meeting Tool

• Input: The meeting transcript.
• AI Step: An AI step with a prompt instructs the language model to summarize the key points from the transcript.
• Slack Integration: The tool sends the summary to a specified Slack channel.

4. Google Docs Template Personalization

• Double Curly Brackets: Double curly brackets are used in the Google Docs template to define variables.
• Create a Document from a Template Module: The "Create a Document from a Template" module in Make.com is used to personalize the proposal.
• Variable Mapping: The variables defined in the Google Docs template are mapped to the data received from the webhook.

5. Stripe Invoice Generation in Make.com

• Create a Customer Module: A "Create a Customer" module is used to create a customer in Stripe.
• Create an Invoice Item Module: A "Create an Invoice Item" module is used to create an invoice item.
• Create an Invoice Module: A "Create an Invoice" module is used to create the invoice, linking it to the customer and invoice item.
• Finalize a Draft Invoice Module: A "Finalize a Draft Invoice" module is used to finalize the invoice.
• Webhook Response: The hosted invoice URL is sent back to Relevance AI via a webhook response.

Importing and Customizing the Agent Template

• Cloning the Agent: The agent can be cloned into a Relevance AI project using a provided link.
• Connecting Software Accounts: The user needs to connect their own Gmail, Google Docs, and Stripe accounts.
• Updating Webhooks: The webhook URLs in the Google Docs Proposal Generator and Stripe Invoice Generator tools need to be updated with the user's own Make.com webhook URLs.

Key Arguments and Perspectives

• Efficiency and Automation: The AI agent significantly reduces the time and effort required to generate proposals and invoices.
• Personalization: The agent creates personalized proposals tailored to each client's specific needs.
• Human Oversight: The human approval step ensures accuracy and allows for adjustments before sending the proposal.
• No-Code Development: The system can be built without coding, making it accessible to a wider audience.
• Integration Capabilities: The combination of Relevance AI and Make.com allows for seamless integration with various third-party services.

Notable Quotes

• "I've been building AI agents for companies since 2023." - Ben
• "...relevance AI is a no code AI agent Builder and uh make.com is a more traditional workflow automation tool..."
• "Always use the escalator manager tool before creating the proposal."

Technical Terms and Concepts

• No-code: Development approach that allows building applications without writing code.
• AI Agent: An autonomous entity that perceives its environment through sensors and acts upon that environment through actuators.
• Webhook: An automated HTTP callback triggered when an event occurs.
• API: Application Programming Interface, a set of rules and specifications that software programs can follow to communicate with each other.
• JSON: JavaScript Object Notation, a lightweight data-interchange format.
• GPT-4o: A large language model created by OpenAI.
• Prompt Engineering: The process of designing and refining prompts to elicit desired responses from language models.
• SOP (Standard Operating Procedure): A set of step-by-step instructions compiled by an organization to help workers carry out complex routine operations.

Logical Connections

The video logically connects the different components of the system, starting with the trigger (Fireflies.ai) and flowing through the Relevance AI agent, Make.com integrations, and finally, the output (personalized proposal and invoice). The video clearly explains how each tool and module contributes to the overall automation process.

Data, Research Findings, or Statistics

The video doesn't explicitly mention specific data, research findings, or statistics. However, it implicitly suggests that the automation saves significant time and effort based on the presenter's experience.

Synthesis/Conclusion

The video provides a comprehensive guide to building a no-code AI agent that automates the generation of personalized proposals and invoices from meeting transcripts. By leveraging the capabilities of Relevance AI, Make.com, and Fireflies.ai, users can create a powerful system that streamlines their sales process and improves efficiency. The key takeaways are the importance of a well-defined agent prompt, the flexibility of no-code platforms, and the power of integrating different services to achieve complex automation.