Skip to main content

Overview: Getting Started with ai12z

This walks you through the process step by step—from creating an account to deploying a fully functional AI assistant. By following these steps, you’ll be able to integrate your copilot into your digital ecosystem, train it with your unique content, and ensure it delivers accurate and valuable insights to your users.

With the flexibility to connect external systems, customize responses, and utilize advanced tools like the ReAct model, your copilot will be a dynamic and powerful solution designed to grow alongside your business. Let’s get started!

1. Sign Up and Create Your Account​

  • Why This Matters: Creating an account is your entry point into the LLM platform. It provides you access to all the tools and settings needed to build your copilot.
  • Detailed Steps:
    • Visit the ai12z website (e.g., ai12z portal). Sign Up
    • Click the "Sign Up" button prominently displayed on the homepage.
    • Fill out the required details:
      • Name: Your full name or your organization’s representative.
      • Email Address: Ensure it’s valid and accessible for verification.
      • Password: Create a secure password following the platform’s guidelines.
      • Organization Details: If applicable, provide company name, size, or industry.
    • Verify your email by clicking the link sent to your inbox.
    • Log in using your new credentials.
  • Outcome: You gain access to the platform’s dashboard and can begin setting up your copilot.

Complete guide for Creating an Account

2. Define Your Copilot’s Use Case​

  • Why This Matters: A well-defined use case helps focus the copilot’s design, ensuring it meets the specific needs of your organization or audience.
  • Detailed Steps:
    • Identify your primary goals:
      • Do you want the copilot to answer customer inquiries, assist in workflows, or provide e-commerce recommendations?
      • Is it internal (employees) or external (customers)?
    • Consider the type of interaction:
      • Search Control, Bot, or a Knowledge box
      • Text-based (chat interface) or image and video.
      • HTML widgets (examples: Morgage Calculator, Google Chart), Carousels: Lists & Sliders
      • Support of Voice commands (speech-to-text and text to speech).
      • Ability for user to upload an image and ask questions
      • Define your Calls to Actions (CTA).
      • Define your goals
    • Type of Copilot
      • Retrieval-Augmented Generation (RAG) is outdated. When comparing solutions ensure the vendor supports ReAct and not RAG only.
      • Modern co-pilots are based on Reasoning and Acting (ReAct) with Agents, Forms, HTML widgets, and RAG as an Agent
        • Out of the box Inegrations, Salesforce, Jira, Office365, Google Maps..
        • Custom Agents integration with your own backend services or others
    • Draft a simple mission statement:
      • Example: "Our copilot will help customers find product information quickly while suggesting relevant accessories."
  • Outcome: A clear roadmap that aligns with your business goals and user expectations.

3. Create a Project​

  • Why This Matters: A project serves as the container for all the configurations, data, and settings specific to a copilot.
  • Detailed Steps:
    • Log in to your account and access the "Projects" section.
    • Click "Create New Project" and provide the following:
      • Name: Something descriptive like "Customer Support Assistant."
      • Organizatin: The Organizatin for who this copilot is for
      • URL of Organization: The Organizatins public website
      • Purpose: A brief summary of the project’s purpose.
      • Language: Select if content being ingested is not english
    • Save and proceed to the next step.
  • Outcome: A dedicated space for building and testing your copilot. Using the correct Organizatin information will be used to populate prompts

Complete guide for Project Creation

4. Add Content for Contextual Understanding​

  • Why This Matters: The quality of the responses depends on the data available. Relevant and structured content ensures the copilot generates accurate, useful answers.
  • Detailed Steps:
    • Website Content:
      • Add your website URL for sitemap ingestion or content crawling.
      • Configure settings to limit ingestion to specific pages or sections.
      • Complete guide to Add a Website
    • Knowledge Base Articles:
      • Navigate to the content upload section.
      • Select "Knowledge Base" and upload or manually enter articles.
      • Use tags and categories for better organization.
      • Complete guide to Add a Knowledge Base Article
    • PDF Documents:
      • Drag and drop important PDF files or browse to upload.
      • Ensure the PDFs are OCR-enabled for easy text extraction.
      • Complete guide to Add a PDF document
    • Advanced Options:
      • Connect directly to your CMS using built-in connectors for real-time updates, as well as other connectors such as Sharepoint.
  • Outcome: A comprehensive repository of your organization's content.

5. Test Asking Questions​

  • Why This Matters: Testing ensures the copilot understands queries and delivers accurate responses, validating its utility before deployment. After creating a project and adding some content. Do your first test, it is that easy.
  • Detailed Steps:
    • Go to the “Test Drive”
    • How to test drive your AI Website Assistant
    • Input various sample questions that mimic real-world user queries.
      • Example: "What are your return policies?" or "How can I schedule a demo?"
    • Evaluate:
      • Relevance of the answers.
      • Response time.
      • Coverage of the content.
    • Refine:
      • If answers are incorrect, adjust content, retrain, or tweak prompts.
  • Outcome: A validated and reliable copilot capable of addressing user needs.

6. Deploy Web Controls Web Page​

  • Why This Matters: Deployment makes your copilot accessible to your audience, enabling real-world interactions.
  • Detailed Steps:
    • Access the deployment section of the platform.
    • Add the web control code snippet:
      • Options for chat, search bar, or interactive knowledge panel.
      • Customize appearance (colors, fonts, branding).
      • Complete guide to adding Web Component into a Web Page
    • Embed the snippet into your website:
      • Add it to your website's HTML, typically before the closing </body> tag.
      • Test the integration on a staging environment.
  • Outcome: A live AI assistant accessible from your website.

7. Categorize Content​

  • Why This Matters: Categories improve the organization of content, making the copilot more efficient at retrieving relevant data.
  • Detailed Steps:
    • Auto generate categories based on your content and the purpose of the Copilot:
      • Business functions (e.g., Sales, Support, FAQs).
      • Content type (e.g., Technical Documents, Tutorials).
    • Assign categories to uploaded content manually or use automatic tagging.
    • Test categorization by enabling "Category Filter" in the Test Drive section.
    • Adjust relevance scores to control how strict the filtering is.
    • See the complete guild to Category Setup
  • Outcome: Users receive more focused and relevant responses.

8. Enhancing ReAct - Reasoning and Acting​

  • Why This Matters: ReAct allows your copilot to combine reasoning with action, making it more versatile and capable. That is the ReAct LLM knows about all the out of the box agents, custom agents and Forms available to use as a tool to accomplish an Action. The reasoning engine has the history of everything it has done for this sessione.
  • Detailed Steps:
    • ReAct System Promt, is created when creating a project, the properties of the Project Creation dialog, specifically the copilot purpose, the Organization name and the Organization URL are used to first create Project Organization Information. Next from this information the AI creates all the prompts.
    • Customize the system prompt:
      • AI creates the system prompt. You should update for your particular use case, for example to add Calls To Actions.
    • Enable necessary agents/tools for your copilot
    • Understanding the workflow
    • Add Custom Agents
  • Outcome: An intelligent copilot capable of not just answering but acting on queries.

9. Set Up Forms​

  • Why This Matters: Forms allow you to collect structured data, which is essential for workflows like lead generation or service requests.
  • Detailed Steps:
    • Use the form builder to design custom forms.
    • Add dynamic controls like dropdowns or date pickers linked to APIs.
    • Integrate forms into the web control for seamless interaction.
    • Optional see how to create a form dynamically as a custom agent
  • Outcome: A richer, more interactive user experience.

10. Set Up Image AI​

  • Why This Matters: Image AI adds visual context to your copilot, enhancing user interactions with relevant images and descriptions.

  • Detailed Steps:

    • Enable "Image Match AI" When answer is created, it will attempt to match an image to the Answer
    • Enable "Image Description AI" When content is ingested, Vision AI will analyze the image, page content and alt text and recreate the alt text.
    • Select image quality (Low, Auto, High) based on your needs.
    • Complete Guide for Setting Up Image AI

  • Outcome: A copilot experience enriched with text and images for more engaging interactions.

11. Test, Monitor, and Optimize​

  • Why This Matters: Continuous testing and optimization ensure the copilot remains effective as user needs evolve.
  • Detailed Steps:
    • Use analytics to monitor user interactions.
    • Review Data tab, - see questions asked, user feedback to identify areas of improvement.
    • Review logs - Validate that the workflow is what is expected.
    • Regularly update content, categories, and configurations.
  • Outcome: A continuously improving copilot that meets user expectations.

12. Create Custom Agents​

  • Why This Matters: Custom Agents let you connect to your digital systems, giving the copilot context, also enables client user controls such as Carousel slider or list, Template HTML widgets, even the ability to process source data with Python.
  • Detailed Steps:
    • Create a Custom Agents - Name and Description, so the LLM knows when to call it.
    • Select Data Source: None, RestAPI, GraphQL
    • Create the LLM parameters
    • Test the endpoint in the Agent
    • Select how to Handle the data response: Send to the LLM, or send the client with a template to visualize the data, such as a Carousel or Template (HTML widget) or Form
    • Ability to use JSONata, both for posting data to an endpoint, and filterting data from a source
    • Enable applying Custom Python code to process data
  • Outcome:

By breaking down each step into detailed actions, you ensure a clear understanding of the process, reducing errors and streamlining implementation.