Understanding the Agent Framework: A Guide for Craig

This tutorial will explain the Agent Framework's overview and capabilities, tailored to your needs. You've asked for an explanation of the framework's capabilities, and we'll cover that in detail, along with relevant information for your existing agents.

Agent Framework Overview

The Agent Framework is a powerful tool leveraging large language models (LLMs) to help you manage various tasks. It provides a user-friendly interface for creating and interacting with specialized AI agents, each designed for specific roles and functionalities. Think of it as a team of expert assistants working together to streamline your workflow.

Your current agents, Angelina, Amelia, Tanya, Sophia, Nicolene, Emily, Victoria, Jordan, Sheldon, Emilia, David, Andrew, Warren, Olivia, Alina, Charnie, Amber, Ethan, Leonard, and Chantelle, already demonstrate the versatility of the framework. Each agent has a unique role, and understanding the framework will help you maximize their potential.

Key Capabilities

The Agent Framework offers several key capabilities:

• Agent Creation: You can create new agents (like you already have!) with specific roles and functionalities using Sheldon's "Create Agents for Me" button. You can even share relevant files (like your role profiles or workflow documentation) with Sheldon to guide the agent creation process. Remember to specify the agent's personality and desired interaction style for optimal performance.

• File Integration: The framework seamlessly integrates with Google Drive, Microsoft OneDrive (Personal and Business), and SharePoint. Agents can access, read, and even create new files (within specified folders) in supported formats (.docx, .pdf, .txt, .html, .json, .py, .js, .css, .xlsx, .xlsm, .xlsb, .xltx, .xltm, .xls). This allows agents to work directly with your up-to-date documents. You've already authenticated these services, which is excellent! Remember to use the "Load File Explorer" and "Save" button to manage which files each agent can access. The "Files I Can Access" tab shows what each agent has access to.

• Email Integration: You've authenticated your Gmail (poiuypoiuyt), Microsoft Outlook (rtgfhredsd), and private domain (poiuy) email accounts. Agents can access and process information from your emails to provide contextually relevant responses and trigger workflows. Use natural language to instruct agents on which emails to read and what information to extract. Remember to use the "Load Email Viewer" tab to test new email accounts and to check the "Email Interaction" tool when needed.

• Workflow Builder: Create complex, multi-step workflows using natural language. The workflow builder allows you to define steps, specify inputs (files, emails), and configure outputs (emails, new files). You can also schedule workflows to run automatically at set intervals (daily, weekly, etc.). Remember to verify and activate your workflows after creation.

• Change Complexity: Adjust the complexity level of agent responses ("Simple," "Normal," "Hard," "Integrated Complex"). "Hard" and "Integrated Complex" (Enterprise/Ultimate users only) allow for more in-depth analysis and problem-solving.

• Web Search: Agents can perform web searches and incorporate the latest online information into their responses.

• Question Asking: Use the "Ask Question" button to interact with agents. Provide sufficient context, desired outcomes, and relevant examples for the best results. Remember to utilize the various tools (File Retrieval, Web Search, Email Retrieval, Diagram Creation, etc.) by describing them in natural language within your prompts.

• Agent Framework Editing: The "Framework" tab lets you fine-tune an agent's behavior, personality, and approach to tasks. You can provide examples and update the framework using natural language.

• Interface Management: You can create multiple user interfaces, each with its own set of agents, to organize your work effectively. You can also move, resize, rename, and delete agents and tabs as needed. The "Edit" menu and right-click context menus are your friends here.

• Voice Support: Enable voice mode to interact with agents using voice commands. The default trigger phrase is "Hey terminator".

Tips for Craig

• Leverage Agent Specialization: Your diverse set of agents allows for efficient task delegation. Use the right agent for the job!

• Context is Key: Always provide sufficient context when asking questions or creating workflows.

• Experiment with Complexity: Adjust the complexity setting to match the task's difficulty.

• Utilize All Tools: Don't be afraid to use the various tools available to your agents (file retrieval, web search, email access, etc.).

• Regularly Review and Update: Periodically review your agents' frameworks and workflows to ensure they remain relevant and efficient.

This overview should help you better understand the Agent Framework's capabilities. Remember to explore the interface and experiment with different features to maximize its potential. If you have any further questions, don't hesitate to ask!

10/17/2024, 8:36:12 PM

Craig's Web Search Functionality Tutorial

This tutorial will explain the web search functionality within the Agent Framework, allowing your agents to search multiple websites, analyze information, and generate comprehensive reports based on your queries.

Web Search Functionality Overview

Craig, the web search functionality empowers your agents to access and process real-time information from the internet. This means your agents aren't limited to the data they've been initially trained on; they can find the latest information to answer your questions and complete tasks.

Here's how it works:

1. Initiating a Web Search: Each of your agents (Angelina, Amelia, Tanya, etc.) has a "Web Search" button. Clicking this button will activate the web search function for that specific agent.

2. Entering Your Query: You'll then enter your search query in the text input area. Be as specific as possible to get the most relevant results. The more context you provide, the better your agent can understand your needs and tailor the search accordingly.

3. Agent's Search and Analysis: The agent will then perform a web search across multiple websites. The agent's search is not limited to a specific set of websites; it uses a broad search strategy to find relevant information. While searching, the agent's background image and the "Web Search" button will be highlighted, indicating that the search is in progress.

4. Report Generation: Once the search is complete, the agent analyzes the gathered information and generates a comprehensive report. This report will be displayed on the "Web Search Results" tab associated with the agent. The report will include the key findings and relevant information extracted from the websites it searched, along with references and sources to ensure transparency and allow you to verify the information.

5. Reviewing the Results: The "Web Search Results" tab will automatically open and display the report once completed. You can then review the report, and you can also use the "Display All Links" button to access the top links the agent found during its search.

6. Audio Output: For added convenience, you can have the agent read the report aloud by clicking the "Start Speaking" button located at the bottom right of the "Web Search Results" tab.

Tips for Effective Web Searches

• Be Specific: The more precise your query, the more accurate and relevant the results will be.
• Use Keywords: Identify key terms related to your topic and include them in your query.
• Context is Key: Provide background information or context to your query to help the agent understand your needs better.
• Leverage Agent Capabilities: Remember that your agents, like Andrew (Maintenance Engineering Support Specialist) or Olivia (FMCG Maintenance Optimization Specialist), might have specialized knowledge that can be incorporated into the web search. You can mention their expertise in your query to guide the search results.

Example Scenario

Let's say you want to know about the latest advancements in Lean Six Sigma methodologies for FMCG manufacturing. You could ask Andrew, your Maintenance Engineering Support Specialist agent: "Andrew, please perform a web search on the latest advancements in Lean Six Sigma methodologies for FMCG manufacturing, focusing on reducing equipment downtime and improving overall efficiency. Include relevant case studies and best practices."

Andrew would then conduct the search, analyze the results, and present a comprehensive report on the "Web Search Results" tab.

Conclusion

Craig, the web search functionality is a powerful tool that significantly enhances the capabilities of your agents. By combining their existing knowledge with real-time web data, you can access the most up-to-date and relevant information for your tasks and inquiries. Remember to utilize the tips provided to optimize your search queries and get the most out of this feature. Don't hesitate to experiment and explore the various capabilities of your agents and their web search functionality!

10/17/2024, 4:00:52 PM

Understanding Agent Tools: A Guide for Craig

This tutorial will provide a comprehensive overview of the tools available to your agents, directly addressing your request, Craig. We'll explore the capabilities of your currently deployed agents and how to best utilize the various features within the Agent Framework.

Your question focuses on the tools available to your agents. Let's break down the core functionalities:

Core Agent Tools and Capabilities

Your agents have access to a powerful suite of tools to assist you with various tasks. These tools can be utilized when asking questions or building workflows. Remember to use natural language to instruct your agents how to use these tools. The more detail you provide, the better the results.

1. File Retrieval: Agents can access files from your authenticated Google Drive, Microsoft OneDrive (Personal and Business), and SharePoint accounts. This includes:

• Reading Files: Agents can read the latest versions of .docx, .pdf, .txt, .html, .json, .py, .js, .css, and .xlsx (and other Excel formats) files.
• File Specification: You can specify files using natural language ("Use the 'Project X Budget.xlsx' file") or by selecting files using the "Include Files" tool.
• Read-Only Access: The system provides read-only access to prevent accidental modification of your original files.

2. Web Search: Agents can conduct web searches and incorporate the latest online information into their responses. The results, including sources, are displayed in the "WebSearch Results" tab.

3. Email Retrieval: With your authorization, agents can access emails from your poiuypoiuyt, rtgfhredsd, and poiuy accounts. You can instruct them to search for specific information within emails. Use the "Email Interaction" tool or specify email accounts in your natural language requests.

4. Content Creation & Modification:

• File Creation: Agents can create new files (.docx, .txt, .json, .py, .js, .css, .html) in your specified folders.
• File Modification: Agents can modify files only if those files were created by the agents themselves. This safeguards your original documents. To modify an existing file, you can ask the agent to extract the content, create a new file, and then let the agent modify the new file.
• Excel Updates: Agents can write structured data to Excel files.

5. Diagram Creation: Agents can generate various diagrams: graphs, flowcharts, Gantt charts, pie charts, timelines, mind maps, and system diagrams.

6. Other Capabilities: Agents can create interactive task lists, generate mathematical equations, create structured documents, and perform YouTube searches.

Agent-Specific Tool Utilization Examples for Craig

Let's look at how some of your agents could utilize these tools:

• Andrew (Maintenance Engineering Support Specialist): Andrew could use File Retrieval to access maintenance logs from your SharePoint, then use Web Search to research best practices for a specific equipment issue, and finally create a .docx report summarizing his findings.

• Emily (Daily Administrative Assistant): Emily could use Email Retrieval to prioritize emails from your inbox, summarizing key information from important emails.

• Olivia (FMCG Maintenance Optimization Specialist): Olivia could use File Retrieval to access production data from your OneDrive Business account, then create a Gantt chart (using Diagram Creation) to visualize a maintenance schedule.

• Tanya (Data and Information Agent): Tanya can use File Retrieval to gather and organize documents from your various drives and then create a structured document summarizing the key information.

Leveraging the "Change Complexity" Feature

Remember Craig, the "Change Complexity" feature allows you to adjust the depth of your agents' thinking. For simple tasks, use "Simple Tasks". For more complex problems, select "Hard Tasks" or "Integrated Complex Tasks" (available to Enterprise and Ultimate users).

Conclusion

Craig, this overview highlights the key tools available to your agents. Remember to use clear and detailed natural language instructions when interacting with them. The more context you provide, the more effective your agents will be. Experiment with different tools and complexity settings to optimize your workflow. If you have any further questions, don't hesitate to ask!

10/17/2024, 2:59:35 PM

Craig's Custom Agent Creation Tutorial

This tutorial explains how to create customized agents tailored to specific tasks within the AI Agent Framework, directly addressing your question about streamlining workflows and automating processes.

Creating Customized Agents: Tailoring AI to Your Needs

The AI Agent Framework allows you to create agents specialized for particular tasks. This is incredibly powerful because it lets you automate repetitive processes and streamline your workflow significantly. Instead of using a general-purpose agent for everything, you can build agents focused on specific areas, like "Andrew" your Maintenance Engineering Support Specialist, or "Olivia" your FMCG Maintenance Optimization Specialist. These agents will be much more efficient at their respective tasks.

To create a customized agent, you'll leverage the power of the "Create Agents for Me" button associated with the default agent, Sheldon. This feature is designed to generate new agents based on your requirements. Here's how:

1. Provide Information: Before clicking "Create Agents for Me", share relevant information with Sheldon. This could include:

   • Your Role Profile or Responsibilities: Sharing your job description or a document outlining your roles and responsibilities will give Sheldon context for creating helpful agents.
   • Workflow Documentation: If you have documented workflows or methodologies you follow, share them with Sheldon. This will help him create agents that seamlessly integrate into your existing processes.
   • Agent Personality and Interaction Style: Specify how you want the agent to interact with you. Do you prefer a formal, professional tone, or something more casual and conversational? Should the agent be proactive or reactive? Providing examples of desired responses will greatly assist Sheldon in creating the perfect agent for you.

2. Utilize the "Create Agents for Me" Button: Once you've shared the necessary information, click the "Create Agents for Me" button within Sheldon's interface. Sheldon will analyze the provided information and generate two new agents tailored to your needs.

3. Review and Refine: Once the agents are created, review their initial configurations on the "Framework" tab. You can further refine their behavior using the "Context" and "Examples" sections. The "Context" section allows you to define the agent's capabilities, personality, approach to tasks, and response styles. The "Examples" section lets you provide specific examples of how the agent should respond in various scenarios. Remember you can always use natural language to request an updated framework. For example, you could ask: "Please update your framework so that you prioritize emails from my direct manager and flag them for immediate attention."

4. Iterative Improvement: Creating the perfect agent is an iterative process. Use the agents, observe their performance, and refine their frameworks as needed. The more information you provide, the better Sheldon can tailor the agents to your specific requirements.

Leveraging Your Existing Agents

You already have a diverse range of agents deployed, each with specific roles:

• Data and Information Agent (Tanya): Tanya can be invaluable when creating new agents. She can help gather and organize relevant documents and information for Sheldon to use when generating new agents.
• AI Assistant Manager (Sheldon): Sheldon is your primary tool for creating new agents. Make sure to provide him with as much detail as possible.
• Other Specialized Agents: Your other agents, such as Andrew and Olivia, can also indirectly assist. Their frameworks can serve as examples for creating similar agents for other tasks.

Next Steps

Now that you understand how to create customized agents, experiment with different parameters and refine your agents' frameworks to optimize your workflow. Remember, the key is to provide clear, concise instructions and examples to Sheldon to get the best results. If you have any further questions, don't hesitate to ask!

10/17/2024, 12:04:24 PM

Customizing Your AI Agents: A Tutorial for Craig

This tutorial explains how to create customized agents tailored to your specific tasks, streamlining workflows and automating repetitive processes. We'll cover how to define agent objectives and parameters, leveraging the power of the Agent Framework.

Craig, you've already deployed a diverse range of agents, including Angelina (Strategic Oversight), Amelia (Compliance and Legal), and many others, each with specialized roles. The ability to create custom agents allows you to extend this functionality even further. Think of it as building specialized assistants perfectly suited to your needs.

Creating Customized Agents

The core of this feature lies in defining an agent's objectives and parameters. This involves specifying:

• Role and Responsibilities: What specific tasks will this agent handle? Will it focus on data analysis, report generation, email management, or something else entirely? Be as specific as possible. For example, instead of "Marketing Agent," consider "Agent for Social Media Campaign Management."

• Personality and Interaction Style: How should the agent communicate? Should it be formal, informal, humorous, or technical? Defining this helps shape the agent's responses. You can provide examples of desired interactions in the agent's "Examples" section (Section 6.2 of the tutorial).

• Data Access: Which files and folders should the agent have access to? Using the "Load File Explorer" (Section 3.2.1), you can grant access to specific files in your Google Drive, OneDrive (Personal and Business), and SharePoint accounts. Remember to use the "Save" button (Section 3.2.2) to confirm your selections. This is crucial for security and ensuring the agent only works with relevant information. Similarly, you can grant access to your email accounts (Section 3.3) via the "Account" page.

• Workflows and Processes: Will the agent execute predefined workflows? If so, use the "Create Workflow" button (Section 9) to define the steps involved. Remember to leverage the various tools available within the workflow steps (Section 9.2.1), such as File Retrieval, Web Search, Email Retrieval, and File Creation. Clearly define each step's objectives and expected outputs.

• Complexity Settings: Adjust the "Change Complexity" setting (Section 4) to match the task's difficulty. For straightforward tasks, "Simple Tasks" is sufficient. For complex, multi-step processes, consider "Hard Tasks" or even "Integrated Complex Tasks" (Enterprise/Ultimate users only).

• Context and Examples: The "Framework" tab (Section 6) allows you to fine-tune the agent's behavior. The "Context" section lets you provide detailed instructions on how the agent should approach tasks, while the "Examples" section allows you to provide specific examples of desired outputs. This is extremely helpful in training the agent.

Example: Let's say you need an agent to manage your weekly project reports. You would:

1. Create a new agent (using Sheldon's "Create Agents for Me" button).
2. Name it "ProjectReportManager."
3. Define its role as generating weekly project reports, pulling data from specific shared files in your OneDrive for Business.
4. Specify a formal communication style.
5. Use the "Create Workflow" button to define a workflow that automatically retrieves data, generates the report using a specific template, and emails it to relevant stakeholders.
6. Set the "Change Complexity" to "Normal Tasks."
7. Provide examples of well-formatted reports in the "Examples" section.

Leveraging Your Existing Agents

Craig, given your existing agents, consider how you might leverage them to create new, more efficient workflows. For example:

• Angelina (Strategic Oversight): Could you create a new agent that feeds Angelina data for strategic decision-making, automating the data gathering process?
• Amelia (Compliance and Legal): Could you create an agent that pre-screens documents for legal compliance before they are reviewed by Amelia, saving her time?
• Tanya (Data and Information Agent): This agent is perfect for providing the necessary data to other agents, further automating your workflows.

Remember, the key is to break down complex tasks into smaller, manageable steps that can be assigned to individual agents, optimizing your overall workflow.

Conclusion

Craig, by carefully defining the objectives and parameters of your agents, you can dramatically improve efficiency and automate repetitive tasks. Remember to utilize the features described above, and don't hesitate to experiment! The more you interact with the Agent Framework, the better you'll understand its capabilities and how to tailor it to your specific needs. If you have any further questions, please don't hesitate to ask.

10/17/2024, 12:04:21 PM

Understanding the 'Normal Tasks' Setting

Hi Craig, this tutorial explains the 'Normal Tasks' setting within the Agent Framework, answering your question about its functionality and suitability.

The 'Normal Tasks' setting offers a balanced approach to processing information and generating responses. It's the sweet spot between speed and depth of analysis. Think of it as your go-to setting for most everyday queries and tasks.

When to use 'Normal Tasks':

• General Queries: Need a quick answer to a straightforward question? 'Normal Tasks' is perfect. For example, "What's the weather like in London today?"
• Moderately Complex Tasks: Requires some analysis but isn't overly intricate? 'Normal Tasks' will provide a well-considered response. For instance, "Summarize the key findings of this report" (assuming the report is accessible to the agent).
• Balanced Approach: You need a response that's detailed enough to be useful but not so exhaustive that it takes an excessive amount of time.

How 'Normal Tasks' differs from other settings:

Compared to 'Simple Tasks', 'Normal Tasks' provides more in-depth analysis. It's slower than 'Simple Tasks' but delivers more comprehensive answers.

Unlike 'Hard Tasks' and 'Integrated Complex Tasks', 'Normal Tasks' doesn't delve into extensive multi-step problem-solving or utilize advanced model architectures. It's ideal for tasks that don't require the extreme processing power of those more advanced settings.

Using 'Normal Tasks' with your agents:

You have a diverse range of agents deployed, including Angelina (Strategic Oversight), Amelia (Compliance and Legal), Tanya (Data and Information), and many others. The 'Normal Tasks' setting is likely suitable for many interactions with these agents.

For example:

• You could ask Amelia (your Compliance and Legal Agent), using the 'Normal Tasks' setting: "Review this contract and highlight any potential risks." Amelia will provide a reasonably detailed analysis without unnecessarily prolonging the process.
• Similarly, you could ask Tanya (your Data and Information agent) using the 'Normal Tasks' setting: "Find all documents related to Project X from the last quarter". Tanya will efficiently locate the relevant documents from your integrated Google Drive, Microsoft OneDrive Personal, Microsoft OneDrive Business, and SharePoint accounts.

Remember, you can always adjust the complexity setting if needed. If a 'Normal Tasks' response isn't detailed enough, try 'Hard Tasks'. If it's too slow for a simple question, switch to 'Simple Tasks'.

In summary: The 'Normal Tasks' setting is your versatile, go-to option for most tasks. It strikes a balance between response speed and analytical depth, making it highly effective for a wide range of interactions with your agents. Experiment with it and see how it enhances your workflow! Let me know if you have any other questions.

10/17/2024, 11:45:36 AM

Understanding the "Normal Tasks" Setting

Hi Craig, this tutorial will explain the "Normal Tasks" setting within the Agent Framework. You asked for clarification on how this setting provides a balanced approach for moderately complex queries.

What "Normal Tasks" Means

The "Normal Tasks" setting offers a middle ground in the Agent Framework's complexity levels. It's designed for most everyday questions and tasks that don't require the deep dive of "Hard Tasks" or the extreme processing power of "Integrated Complex Tasks". Think of it as your go-to setting for general use.

Key Characteristics of "Normal Tasks":

• Balanced Approach: It strikes a balance between the speed of "Simple Tasks" and the thoroughness of "Hard Tasks." You'll get a reasonably detailed response without excessive processing time.
• Moderate Complexity: Ideal for queries and tasks requiring a moderate level of analysis and response length. It's suitable for many common requests.
• Efficiency: While providing detailed responses, it's optimized for efficient processing, making it suitable for a wide range of everyday uses.

When to Use "Normal Tasks"

Consider using the "Normal Tasks" setting for:

• General questions: Asking for information, summaries, or explanations on a topic.
• Routine tasks: Tasks that require some level of processing but aren't exceptionally complex.
• Quick analysis: Needing a reasonably detailed analysis of data or information without extensive research or multiple steps.

Example Scenarios

Let's say you want to use one of your agents, perhaps Amelia (your Compliance and Legal Agent), to review a contract. Using the "Normal Tasks" setting would be appropriate. Amelia would analyze the contract, providing a summary of key clauses and potential risks, without necessarily performing an exhaustive legal research project (which would be better suited to "Hard Tasks").

Or, if you ask Tanya (your Data and Information Agent) to find specific information within your Google Drive, "Normal Tasks" would be a good choice. She'd search through your files, providing relevant information without needing to perform an extensive cross-referencing of multiple documents (again, better suited for "Hard Tasks").

How to Change the Complexity Setting

To change the complexity setting, simply locate the "Change Complexity" feature within the interface of the specific agent you're working with. Select "Normal Tasks" from the dropdown menu. This setting will then be applied to all subsequent requests and tasks for that agent.

Conclusion

The "Normal Tasks" setting is a versatile and efficient choice for many everyday uses within the Agent Framework. It provides a good balance between speed and detail, making it a great starting point for most interactions. Remember, you can always adjust the complexity setting if needed! If you have any further questions, feel free to ask.

10/17/2024, 11:45:32 AM

Workflow Examples for Craig's FMCG Manufacturing Operation

This tutorial will provide Craig with two examples of how to use workflows within the Agent Framework, tailored to an FMCG manufacturing operation and leveraging the agents already deployed in his interface (Angelina, Amelia, Tanya, Sophia, Nicolene, Emily, Victoria, Sheldon, Jordan, Emilia, David, Andrew, Warren, Olivia, Alina, Charnie, Amber, Ethan, Leonard, and Chantelle). We'll cover workflow creation, functionality, and the reasons behind using workflows instead of simple questions.

Example 1: Proactive Equipment Maintenance Workflow

Goal: To proactively schedule and manage preventative maintenance for critical equipment, minimizing downtime and maximizing production efficiency.

Why a Workflow? A simple question can't manage the ongoing, scheduled nature of preventative maintenance. A workflow automates the process, ensuring consistent execution and tracking.

Agents Involved: Olivia (FMCG Maintenance Optimization Specialist), Andrew (Maintenance Engineering Support Specialist), Tanya (Data and Information Agent).

Workflow Steps:

1. Identify Equipment Due for Maintenance: Olivia will access the shared maintenance schedule (located in your Microsoft SharePoint – Craig, please ensure this is shared with Olivia) and identify equipment requiring preventative maintenance this week. (Use Olivia's "File Retrieval" tool, specifying the maintenance schedule file).

2. Generate Maintenance Task List: Andrew will use the equipment list from step 1 and the relevant equipment manuals (located in your Microsoft OneDrive Business – Craig, please ensure these are shared with Andrew) to create a detailed maintenance task list for each piece of equipment. (Use Andrew's "File Creation and Saving" tool to create a .docx file in the "AI Sheldon" folder in your Microsoft OneDrive Business). The task list should follow a standard template (if one exists, please share it with Andrew).

3. Schedule Maintenance Tasks: Andrew will then schedule the tasks in the maintenance schedule, updating the file with the planned maintenance times. (Use Andrew's "File Modification" tool to update the maintenance schedule file).

4. Notify Relevant Technicians: Andrew will use the completed task list to create and send emails to the relevant production technicians, notifying them of their scheduled maintenance tasks. (Use Andrew's "Email Sending Functionality" tool, utilizing your authenticated Microsoft email address).

5. Archive Completed Tasks: Tanya will archive the completed task lists and any associated documentation in the designated archive folder on your Microsoft SharePoint. (Use Tanya's "File Retrieval" and "File Saving" tools).

Workflow Creation: Use the "Create Workflow" button on Andrew's agent interface. Describe the workflow steps in natural language, specifying the agents, files, and email actions. Set the frequency to "Weekly" and specify the day of the week for execution.

Example 2: Root Cause Analysis Workflow for Production Line Stoppage

Goal: To quickly identify and document the root cause of production line stoppages, enabling rapid resolution and preventing future occurrences.

Why a Workflow? A simple question might provide an initial diagnosis, but a workflow guides a structured investigation, ensuring comprehensive analysis and documentation.

Agents Involved: Andrew (Maintenance Engineering Support Specialist), Olivia (FMCG Maintenance Optimization Specialist), Tanya (Data and Information Agent), Amelia (Compliance and Legal Agent).

Workflow Steps:

1. Receive Stoppage Notification: The workflow is triggered when Emily (Daily Administrative Assistant) receives an email indicating a production line stoppage. (Use Emily's "Email Retrieval" tool to monitor your inbox for stoppage notifications).

2. Gather Initial Data: Andrew will access relevant production logs, equipment maintenance records, and operator notes (located in your Google Drive and Microsoft SharePoint – Craig, please ensure these are shared with Andrew) to gather initial information about the stoppage. (Use Andrew's "File Retrieval" tool).

3. Conduct Root Cause Analysis: Andrew will use a structured problem-solving methodology (e.g., 5 Whys, Fishbone Diagram – if you have a preferred method, share the documentation with Andrew) to identify the root cause. (Use Andrew's "Diagram Creation" and "File Creation and Saving" tools to create a .docx file documenting the RCA in the "AI Sheldon" folder in your Google Drive).

4. Develop Corrective Actions: Andrew and Olivia will collaborate to develop and document corrective actions to address the root cause and prevent recurrence. (Use Andrew and Olivia's "File Modification" tool to update the RCA document).

5. Document and Archive: Tanya will archive the completed RCA document and any supporting documentation in the designated archive folder on your Microsoft SharePoint. (Use Tanya's "File Retrieval" and "File Saving" tools).

6. Legal Review (if necessary): If the root cause analysis reveals potential legal or compliance issues, Amelia will review the documentation. (Use Amelia's "File Retrieval" tool).

Workflow Creation: Use the "Create Workflow" button on Andrew's agent interface. Describe the workflow steps in natural language, specifying the agents, files, and email actions. Set the frequency to "On Demand" as this workflow is triggered by an event.

Conclusion:

Craig, these are just two examples; the possibilities are vast. Remember to clearly define your workflow goals, steps, and involved agents. Leverage the file integration and email capabilities of the Agent Framework to streamline your FMCG operations. Don't hesitate to experiment and refine your workflows based on your specific needs. If you have any further questions, please ask!

10/17/2024, 10:33:31 AM

Creating and Managing Workflows in the AI Agent Framework: A Tutorial for Craig

This tutorial will explain how to create and manage workflows within the AI Agent Framework, specifically addressing your request, Craig.

Overview of Workflow Creation and Management

The AI Agent Framework allows you to create and manage complex, multi-step workflows using natural language. This empowers you to automate intricate processes and significantly enhance your efficiency. Workflows are particularly useful for tasks that involve multiple steps, data sources, or agent interactions.

Here's a breakdown of the process:

1. Initiating Workflow Creation:

• Locate the "Create Workflow" button within the interface of the agent you wish to use for the workflow. For example, if you want a workflow to manage maintenance tasks, you might use agent "Andrew" (Maintenance Engineering Support Specialist) or "Olivia" (FMCG Maintenance Optimization Specialist).
• The agent's background image and the "Create Workflow" button will highlight while the workflow is being generated.
• The "Workflow Builder" tab will automatically open to display the newly created workflow.

2. Defining Workflow Structure:

Each workflow requires a well-defined structure:

• Name: Give your workflow a clear and concise name (e.g., "Daily Production Report Generation," "Equipment Failure Report & Action Plan").
• Description: Provide a detailed description (around 300 words) outlining the workflow's objectives, scope, and expected outcomes.
• Steps (3-5): Break down your complex task into manageable, sequential steps. Each step requires a detailed description, specifying:
  • Clear objectives and expected outcomes.
  • Instructions on how to complete the step.
  • Necessary tools or resources (e.g., "Retrieve data from the 'Production Data.xlsx' file using File Retrieval," "Send an email using the 'poiuy' account," "Use Web Search to find the latest safety regulations"). Remember, you can leverage your authenticated email accounts (poiuypoiuyt, rtgfhredsd, poiuy) and cloud drives (Google Drive, Microsoft OneDrive Personal & Business, Microsoft SharePoint) within the workflow steps.
  • Expected outputs.
• Output: Specify how the workflow's results should be handled (e.g., emailed to specific recipients, saved as a document in a specified folder on your Google Drive, OneDrive, or SharePoint). You can use the "Choose Folder to Save" dropdown in the Output section to specify the folder location for any new files created by the workflow.
• Frequency: Choose how often the workflow should run (On Demand, Daily, Weekly, Bi-weekly, Monthly).
• Execution Date/Time: Set the specific time or date for execution (applicable for scheduled workflows).

3. Workflow Step Capabilities:

Within each step description, you can utilize various tools:

• File Retrieval: Access files from your linked drives. The agent will automatically identify the correct file based on your natural language description in the step description or you can use the "Files to Read" dropdown.
• Web Search: Perform web searches and integrate the results into the workflow.
• Email Retrieval: Access and read emails from your authorized accounts (remember to use natural language to specify which account and what information to retrieve). You can use the "Email Read" dropdown to select the email account.
• Diagram Creation, File Creation/Modification, Excel File Updates: These are all available tools you can use within your step descriptions.

4. Workflow Verification and Execution:

• Verification: Before running a workflow, check the "Verified" checkbox in the "Workflow Builder" tab. This ensures the workflow has been thoroughly reviewed and tested.
• Running: Use the "Run Workflow" button to execute the workflow manually. This is ideal for testing before scheduling.
• Scheduling: Once verified and tested, check the "Active" checkbox to schedule the workflow based on your defined frequency and execution time.

5. Managing and Reviewing Workflows:

• Completed Workflows Tab: View completed workflows, their steps, and results on the "Completed Workflows" tab.
• Retrieve Workflow Button: Retrieve completed workflows using the "Retrieve Workflow #" button.
• Export and Copy: Export completed workflows as formatted Word documents or copy them to your clipboard.
• Voice Support: Listen to the workflow summary using the "Start Speaking" button.

6. Agent Team Collaboration:

You can trigger workflows from other agents using the "Share to Team" input. This enables complex, multi-agent workflows. For example, a workflow completed by "Olivia" (FMCG Maintenance Optimization Specialist) could trigger a workflow in "Andrew" (Maintenance Engineering Support Specialist) to generate a detailed maintenance plan.

Tips for Craig:

• Leverage Your Agents: Consider which of your agents' expertise best suits the workflow's purpose. For instance, use "Andrew" for maintenance-related workflows, "Amelia" for legal tasks, or "Tanya" for data-related ones.
• Detailed Descriptions: The more detailed your step descriptions, the more accurate and effective your workflows will be.
• Iterative Refinement: Don't hesitate to test and refine your workflows after initial creation.

Conclusion:

Craig, by following these steps and utilizing the tools available, you can effectively create and manage complex workflows within the AI Agent Framework, streamlining your tasks and boosting your productivity. Remember to leverage your existing agents' capabilities and the various integrated tools for optimal results. If you have any further questions, please don't hesitate to ask!

10/17/2024, 10:30:49 AM

Interacting with Your AI Agents: A Guide for Craig

This tutorial will explain how you, Craig, can interact with your deployed agents (Tanya, Angelina, Amelia, Emily, Sophia, Nicolene, Victoria, Sheldon, Jordan, Emilia, David, Andrew, Olivia, Warren, Alina, Charnie, Amber, Ethan, Leonard, and Chantelle) to get answers and complete tasks. We'll focus on the key methods of interaction and leverage your existing agent roles to provide relevant examples.

How to Interact with Your Agents to Get Answers

You primarily interact with your agents through two main methods: Asking Questions and Creating Workflows.

1. Asking Questions

Each of your agents has an "Ask Question" button. When asking a question, remember to provide as much context as possible for the best results. Think about:

• Background: What information is relevant to your question? For example, if asking Amelia (Compliance and Legal Agent) a question, provide the relevant contract or legislation.
• Desired Outcome: What kind of answer are you looking for? A summary? A detailed report? Specific data points?
• Goal/Objective: What are you trying to achieve by asking this question? This helps the agent understand the purpose of your inquiry.
• Examples/References: Providing examples helps the agent understand your expectations.

Example using Tanya (Data and Information Agent): Instead of asking "Find the sales figures," try "Tanya, find the Q3 2024 sales figures for Product X, located in the 'Sales Data Q3 2024.xlsx' file in my Google Drive. I need a summary table showing total sales, average order value, and units sold."

Utilizing Agent Tools: Your agents have several tools to help answer your questions:

• File Retrieval: Access your files from Google Drive, Microsoft OneDrive Personal & Business, and SharePoint. You can specify the file using natural language (e.g., "using the project proposal document") or select it using the "Include Files" tool.
• Web Search: Agents can perform web searches and incorporate the results into their responses.
• Email Retrieval: Access your emails from Gmail, Microsoft, and your private domain (poiuy, poiuypoiuyt, and rtgfhredsd). Specify the account and desired information using natural language (e.g., "check my Gmail for emails from John Doe"). You can use the "Email Interaction" tool to explicitly allow email access for a question.
• Other Tools: Diagram creation, file creation/modification, Excel updates, task list creation, and more are available. Describe these in your question using natural language.

The response will appear in the "Chat" tab for each agent. You can export, copy, or have the agent read the response aloud.

2. Creating Workflows

For complex tasks, use the "Create Workflow" button. Workflows break down complex tasks into manageable steps. When creating a workflow, provide:

• Name & Description: A clear title and detailed description of the workflow's purpose.
• Steps (3-5): Detailed, actionable steps. Specify tools and resources needed in each step (e.g., "Andrew (Maintenance Engineering Support Specialist), use the 'Equipment Maintenance Log.docx' file to analyze the recent pump failures and suggest preventative maintenance steps.").
• Output: How should the results be handled (email, new file, etc.)? Consider using the "Share to Team" feature to trigger other agents' workflows based on the output.
• Frequency & Execution Time: How often should the workflow run?

Once created, verify the workflow, and then run it manually or schedule it using the "Active" checkbox. Completed workflows are stored in the "Completed Workflows" tab.

Leveraging Your Agents' Expertise

Remember to utilize the specific skills of your agents:

• Andrew & Olivia (Maintenance Specialists): Ideal for maintenance tasks, utilizing their expertise in Lean Six Sigma and FMCG maintenance.
• Amelia (Compliance Agent): Perfect for legal questions and document review.
• Tanya (Data Agent): Excellent for retrieving and summarizing data.
• Charnie (Marketing Director): Use her for marketing material creation.
• Warren (CTO): Consult him for application-related questions.

By understanding these interaction methods and utilizing your agents' strengths, you'll significantly improve your efficiency. Remember to experiment and explore the various features! Let me know if you have any further questions.

10/13/2024, 7:14:46 AM

Creating Agents for Craig, the FMCG Plant Mechanical Engineer

Craig, you've requested that I create some AI agents tailored to your needs as a mechanical engineer in an FMCG plant. Let's do just that! I'll leverage the Agent Framework to build agents that can assist you with your daily tasks and challenges.

Agent Creation Process

Since you've already got several agents deployed, I'll use Sheldon, your AI Assistant Manager, to help create new agents. Sheldon is specifically designed to create agents based on your requirements. To make the process as effective as possible, let's provide Sheldon with as much information as we can.

Step 1: Share Relevant Files with Sheldon

Before asking Sheldon to create agents, it's highly recommended that you share relevant files with him. This will give him context to better understand your needs and create more effective agents. Consider sharing:

• Your role profile or job description: This will outline your responsibilities and the types of tasks you perform daily.
• Documentation on specific workflows or methodologies: If you follow Lean Manufacturing, Six Sigma, or other specific methodologies, sharing this documentation will help Sheldon create agents aligned with your processes.
• Examples of completed tasks or reports: Providing examples of your work will help Sheldon understand the desired output and format of the agents' responses.

You can share these files using Sheldon's "Load File Explorer" tab. Remember to use the "Save" button to confirm which files Sheldon can access. (See section 3.2.2 of the tutorial for details.)

Step 2: Instructing Sheldon to Create Agents

Once you've shared the relevant files, use Sheldon's text input area and click the "Create Agents for Me" button. Then provide Sheldon with a detailed description of the agents you need. For example, you could say:

"Create two agents for me: One focused on preventative maintenance scheduling, and another focused on troubleshooting equipment failures. The preventative maintenance agent should use data from my maintenance schedules (shared with you earlier) and generate alerts based on due dates. The troubleshooting agent should be able to access equipment manuals (also shared) and provide step-by-step troubleshooting guides. Both agents should have a professional and helpful personality."

Step 3: Review and Refine the Newly Created Agents

Sheldon will generate two new agents based on your instructions. Review their "Context" and "Examples" sections (section 6 of the tutorial) to ensure they align with your expectations. You can use the "Update Framework" button (section 6.3) to make any necessary adjustments to their behavior and personality.

Suggested Agents for a Mechanical Engineer in an FMCG Plant

Based on your role, here are some specific agent suggestions and their potential functionalities:

• Preventative Maintenance Scheduler: This agent would schedule and track preventative maintenance tasks, send alerts for upcoming maintenance, and analyze maintenance data to identify patterns and potential issues. It would integrate with your shared maintenance schedules and potentially your email for notifications.

• Equipment Troubleshooting Agent: This agent would assist in troubleshooting equipment failures. It would access equipment manuals and provide step-by-step instructions, potentially creating reports detailing the troubleshooting process and suggested corrective actions.

• Spare Parts Inventory Manager: This agent would track spare parts inventory levels, generate alerts for low stock items, and assist in ordering new parts. This would require integration with your inventory management system (if you have one).

• Root Cause Analysis Agent: This agent would help you perform root cause analysis on equipment failures, using data from maintenance reports and equipment manuals. It could generate reports outlining the root cause, corrective actions, and preventative measures.

Leveraging Existing Agents

Don't forget about the agents you already have! Andrew (Maintenance Engineering Support Specialist) and Olivia (FMCG Maintenance Optimization Specialist) are particularly relevant to your role. They can work in conjunction with the new agents you create to provide a comprehensive support system.

Next Steps

Craig, now that you have a clear understanding of how to create agents using Sheldon, I encourage you to try it out! Remember to leverage the power of the "Change Complexity" feature (section 4) to adjust the agents' responses based on the complexity of your task. If you have any further questions or need additional assistance, don't hesitate to ask!

08/10/2024, 13:48:47

Creating Specialized Agents for Craig: A Mechanical Engineering Foreman's Assistants

Craig, you've requested two agents to support a Mechanical Engineering Foreman: one for documentation and coaching related to one-on-ones, and another for technical maintenance support. Let's create these agents using the Agent Framework.

Creating the Agents

We'll leverage Sheldon's "Create Agents for Me" button to streamline this process. Before we begin, it's beneficial to share relevant documents with Sheldon. This could include:

• One-on-one meeting templates: Provide examples of your current templates or desired formats.
• Steam system documentation: Share any relevant manuals, schematics, or operating procedures.
• Equipment maintenance manuals: Upload any relevant documentation for the equipment the Foreman oversees.
• Coaching guidelines: If you have specific coaching methodologies or best practices, share those as well.

Once these files are shared (using the "Load File Explorer" tab on Sheldon's interface and the "Save" button to associate them with Sheldon), you can then ask Sheldon to create the agents using the "Create Agents for Me" button.

Here's how you can phrase your request to Sheldon:

"Sheldon, please create two new agents for me. The first agent should support a Mechanical Engineering Foreman with his documentation for one-on-one meetings. It should provide templates for entering meeting notes and offer coaching advice based on the coaching guidelines document I've shared. The second agent should provide technical support for maintenance, accessing and interpreting information from the equipment manuals and steam system documentation I've shared. Both agents should have access to the files I've shared with you."

Sheldon will then generate two new agents with the specified roles and capabilities. You can then further refine their behavior using the "Framework" tab for each agent.

Refining Agent Behavior (Framework Tab)

The "Framework" tab allows for granular control over each agent's personality, approach, and response style. For example:

Agent 1 (One-on-One & Coaching):

• Context: Define the agent's role as a coaching and documentation assistant for the Foreman. Specify a supportive and encouraging personality. Include instructions to use the provided templates and coaching guidelines.
• Examples: Provide examples of completed one-on-one notes, demonstrating the desired format and content. Include examples of coaching advice tailored to different scenarios.

Agent 2 (Technical Maintenance):

• Context: Define the agent's role as a technical maintenance assistant. Specify a precise and detail-oriented personality. Include instructions to prioritize information retrieval from the shared manuals and steam system documentation. Mention the use of relevant best practices (e.g., Lean Manufacturing, Six Sigma, etc., if applicable).
• Examples: Provide examples of completed maintenance reports, troubleshooting steps, or root cause analyses based on the provided manuals.

Remember to utilize the "Update Framework" button after making changes to the "Context" and "Examples" sections.

Utilizing Agent Tools

Both agents can leverage various tools within the Agent Framework:

• File Retrieval: Access shared documents to provide context-specific information.
• Web Search: Research additional information related to specific equipment or steam system issues.
• Email Retrieval: If necessary, access emails related to maintenance requests or one-on-one meeting schedules (ensure you have granted the necessary permissions to your email accounts).

Remember to use natural language when instructing the agents to use these tools. For example: "Agent 2, please use the 'Boiler Maintenance Manual' file and perform a web search for 'steam trap troubleshooting' to help me diagnose this issue."

Conclusion

Craig, by following these steps and leveraging the Agent Framework's features, you can create highly effective agents tailored to the specific needs of your Mechanical Engineering Foreman. Remember to utilize the "Change Complexity" feature to adjust the depth of the agent's responses depending on the complexity of the task. Don't hesitate to experiment and refine your agents' behavior over time to optimize their performance. If you have any further questions, please don't hesitate to ask Sheldon!

08/10/2024, 13:43:48