--- title: "Exploring Agents & AutoGen: Enhancing Workflow Automation" description: Discover the latest advancements in AutoGen and multi-agent frameworks in the AI field with our expert Tim Matthews, a Kubernetes Platform Engineer. image: https://analytium.com/hubfs/Blog%20Images/July%2024/Exploring%20Agents%20%26%20AutoGen%20Enhancing%20Workflow%20Automation%20.jpg --- [Skip to main content](https://analytium.com/resource-hub/exploring-agents-and-autogen-enhancing-workflow-automation#main-content) [![Analytium Logo](https://analytium.com/hs-fs/hubfs/Logo%20MidRes.png?width=250&height=67&name=Logo%20MidRes.png) ![Analytium Logo](https://analytium.com/hs-fs/hubfs/Logo%20MidRes.png?width=250&height=67&name=Logo%20MidRes.png) ![Analytium Logo](https://analytium.com/hs-fs/hubfs/Logo%20MidRes.png?width=250&height=67&name=Logo%20MidRes.png) ![Analytium Logo](https://analytium.com/hs-fs/hubfs/Logo%20MidRes.png?width=250&height=67&name=Logo%20MidRes.png)](https://analytium.com) - Show submenu for Solutions Solutions - Show submenu for Advice Advice - [Data Maturity](https://analytium.com/en/data-maturity/) - [Data Strategy Consulting](https://analytium.com/en/data-strategy-consulting) - Show submenu for Build and Deploy Build and Deploy - [Agentic AI](https://analytium.com/en/agentic-ai) - [Artificial Intelligence (LLM)](https://analytium.com/en/artificial-intelligence/) - [Business Intelligence and Visualisation](https://analytium.com/en/business-intelligence-and-visualisation) - [Data Integration & Quality](https://analytium.com/en/data-integration-and-quality/) - [Data Science and Predictive Modelling](https://analytium.com/en/data-science-and-predictive-modelling) - [Data Warehouse](https://analytium.com/data-warehouse-implementation-and-deployment/) - Show submenu for Maintain and Optimise Maintain and Optimise - [Data Platform Assessment](https://analytium.com/en/data-platform-assessment/) - [Maintenance and Support](https://analytium.com/en/maintenance-and-support) - [Platform Modernisation](https://analytium.com/en/platform-modernisation) - Show submenu for Workforce Solutions Workforce Solutions - [Flexible Workforce](https://analytium.com/en/flexible-workforce) - Show submenu for Technology Expertise Technology Expertise - Show submenu for Microsoft Microsoft - [Azure Cognitive Services](https://analytium.com/azure-cognitive-services) - [Azure Data Fabric](https://analytium.com/en/azure-fabric) - [Azure Data Platform](https://analytium.com/en/azure-data-platform) - [Azure Machine Learning](https://analytium.com/en/azure-machine-learning) - [Microsoft Power BI](https://analytium.com/en/microsoft-power-bi) - Show submenu for SAS® SAS® - [SAS® Viya and SAS Language](https://analytium.com/en/sas-viya-and-sas-language) - [SAS® Mainframe](https://analytium.com/en/sas-on-mainframe) - [Resource Hub](https://analytium.com/resource-hub) - Show submenu for About Us About Us - [Company](https://analytium.com/en/about-us) - [Careers](https://analytium.com/careers) - [Contact Us](https://analytium.com/en/contact-us) Open main navigation Close main navigation - Show submenu for Solutions Solutions - Solutions - Show submenu for Advice Advice - Advice - [Data Maturity](https://analytium.com/en/data-maturity/) - [Data Strategy Consulting](https://analytium.com/en/data-strategy-consulting) - Show submenu for Build and Deploy Build and Deploy - Build and Deploy - [Agentic AI](https://analytium.com/en/agentic-ai) - [Artificial Intelligence (LLM)](https://analytium.com/en/artificial-intelligence/) - [Business Intelligence and Visualisation](https://analytium.com/en/business-intelligence-and-visualisation) - [Data Integration & Quality](https://analytium.com/en/data-integration-and-quality/) - [Data Science and Predictive Modelling](https://analytium.com/en/data-science-and-predictive-modelling) - [Data Warehouse](https://analytium.com/data-warehouse-implementation-and-deployment/) - Show submenu for Maintain and Optimise Maintain and Optimise - Maintain and Optimise - [Data Platform Assessment](https://analytium.com/en/data-platform-assessment/) - [Maintenance and Support](https://analytium.com/en/maintenance-and-support) - [Platform Modernisation](https://analytium.com/en/platform-modernisation) - Show submenu for Workforce Solutions Workforce Solutions - Workforce Solutions - [Flexible Workforce](https://analytium.com/en/flexible-workforce) - Show submenu for Technology Expertise Technology Expertise - Technology Expertise - Show submenu for Microsoft Microsoft - Microsoft - [Azure Cognitive Services](https://analytium.com/azure-cognitive-services) - [Azure Data Fabric](https://analytium.com/en/azure-fabric) - [Azure Data Platform](https://analytium.com/en/azure-data-platform) - [Azure Machine Learning](https://analytium.com/en/azure-machine-learning) - [Microsoft Power BI](https://analytium.com/en/microsoft-power-bi) - Show submenu for SAS® SAS® - SAS® - [SAS® Viya and SAS Language](https://analytium.com/en/sas-viya-and-sas-language) - [SAS® Mainframe](https://analytium.com/en/sas-on-mainframe) - [Resource Hub](https://analytium.com/resource-hub) - Show submenu for About Us About Us - About Us - [Company](https://analytium.com/en/about-us) - [Careers](https://analytium.com/careers) - [Contact Us](https://analytium.com/en/contact-us) - [Let's Talk](tel:+442035748313) [Let's Talk](tel:+442035748313) Recently, there's been a lot of buzz around AutoGen, especially if you have been following AI newsfeeds and channels. Experts and innovators have ear marked multi-agent frameworks as the potential next leap in the application of AI. So, let's dive into what this is all about.  ## **WHAT ARE AGENTS?** ![Icon Graph metrics surrounded by gear and arrows in circular motion](https://analytium.com/hubfs/Icons/Icon%20Graph%20metrics%20surrounded%20by%20gear%20and%20arrows%20in%20circular%20motion.svg) An Agent is essentially a piece of **Python code that defines its configuration**. Typically, it includes a **source LLM (Large Language Model)**, which could be a **chatbot** or **one fine-tuned with programming examples, and a prompt**.  ![Icon Blue gear icon with circular design and smaller gears around the perimeter.](https://analytium.com/hubfs/Icons/Icon%20Blue%20gear%20icon%20with%20circular%20design%20and%20smaller%20gears%20around%20the%20perimeter..svg) The **prompt provides initial instructions to the Agent**, such as:  ***"You are a project planner. You receive work requests, break them down into small, manageable tasks, monitor the progress of each task, and ensure they are all completed successfully." *** ![Icon Tablet with gear and speech bubbles](https://analytium.com/hubfs/Icons/Icon%20Tablet%20with%20gear%20and%20speech%20bubbles.svg) Agents can be **equipped with various tools**. These tools might include the **ability to execute shell scripts, run Python code, or perform specific actions** like calling a web service that returns structured data. Tools can **help overcome limitations** like outdated training data by executing live internet searches.  ![Icon 3 individuals working together to get a task playing or started](https://analytium.com/hubfs/Icons/Icon%203%20individuals%20working%20together%20to%20get%20a%20task%20playing%20or%20started.svg) An **Agent can receive a message** (from a human or another Agent) and then **perform work**. Agents can **operate individually or in groups**, utilising various orchestration modes. They send and receive messages, **enabling coordination** between the Agent, user, and optionally other Agents.  ## **WHY IS THIS BENEFICIAL?** ![Icon Lightbulb representing an idea with arrow target and notepad](https://analytium.com/hubfs/Icons/Icon%20Lightbulb%20representing%20an%20idea%20with%20arrow%20target%20and%20notepad.svg) ### Challenges and Limits of LLMs No single LLM is perfect. While recent chatbots excel at understanding instructions and newer code models reliably produce usable code, **no one model excels at every task**. This might change over time, but there are **inherent technical challenge**s.  LLM **attention is crucial for good output**, but large context windows are expensive, slow, and have computational limits. Attention computation scales quadratically with context size:  - **If context = 4, then computation = 16 ** - **If context = 6, then computation = 36 ** - **If context = 10, then computation = 100 ** The **unit of computation is referred to as a FLOP**. Each GPU variant can perform a **certain number of FLOPs/second**, which determines its **inference speed**.  ![Icon representing data backup process from a central database to individual folders.](https://analytium.com/hubfs/Icons/Icon%20representing%20data%20backup%20process%20from%20a%20central%20database%20to%20individual%20folders..svg) ### Implementing Agent-Based Architecture for Efficient LLM Usage **To get useful output from LLMs, key questions and data must be within the attention window**. Large context models are available but are **expensive and slower due to the required FLOPs**. Cheaper models might respond faster but have smaller attention windows.  Agents allow us to **split these attention requirements**. For example, we can define a **Planner Agent** using GPT-4o (for better instructions but at a higher cost) and a **Worker Agent** using GPT-4o-mini (cheaper and faster). The **Planner acts as an interface** between the human task and the Worker, **breaking down requests into small, self-contained tasks** and ensuring their successful completion.  In this setup, the Planner might be asked to ***"write code to plot stock market performance for stock ticker symbols provided by the user."*** The Planner could split this into tasks like:  - **Write code to identify the stock ticker symbol for a given company. ** - **Write code to retrieve historical stock prices for a given symbol. ** - **Write code to plot data on a chart. ** - **Test the code using the symbol AAPL for 2024. ** ![Icon representing a central figure with multiple connections or options.](https://analytium.com/hubfs/Icons/Icon%20representing%20a%20central%20figure%20with%20multiple%20connections%20or%20options..svg) **The Worker Agent is more likely to successfully implement each sub-task without hallucinating** because the key requirements will be within its context window during token generation. Additionally, the **Planner Agent tests each task**, sending any issues back to the Worker Agent. **Since each task is small, the Worker Agent can resolve problems with both the code and the Planner's feedback in its context. ** ## **INTRODUCING AUTOGEN** ![Isometric illustration depicting a graph in a funnel, surrounded by digital displays representing an analytics platform](https://analytium.com/hubfs/Graphics/July%2024%20500%20x%20500/SAS%20Analytics%20Platforms%20Flexible%20Workforce%20Solutions.png) **AutoGen** is an **open-source framework from Microsoft for programming and managing agent workflows**. It provides **base classes** to **define Agents** and **their behaviors** and **orchestration methods to control Agent-Agent and Agent-Human interactions. ** In more complex scenarios, we might have a Planner, a Coder, and a Writer (to document the work). **The Planner ensures the Coder completes all tasks before passing the results to the Writer, who documents the entire process.** The Writer may not need to see all previous conversations—just the final code—or might need to know everything that occurred during the previous steps to write a history. **AutoGen provides tools to specify how the conversation flows, what information is pushed forward, and what is retained. ** ![Isometric illustration of a cloud on a platform with bar graphs and screens displaying data](https://analytium.com/hubfs/Graphics/July%2024%20500%20x%20500/Cloud-Based%20Customer%20Experience%20Analytics%20for%20Marketing.png) ## **SUMMARY** This is just a brief overview of Agents and AutoGen. **Analytium is currently working on an Agent Flow to perform feature identification and extraction, to automatically convert configuration data from one platform to another, validate the conversion (providing corrective feedback if necessary), and then test and document the results.** If you are interested, get in touch, and we’ll share more.  ## **INTERESTED IN LEARNING MORE?** ## Book a Free 30-Minute Consultation Your business' data has potential and it can reach new heights with a **Free 30-Minute Consultation** from Analytium! Book a meeting with our expert, **Sander De Hoogh**, and get personalised insights into optimising your data strategies. Thank you for considering Analytium. We look forward to helping you achieve your data-driven goals. Click below to schedule your consultation and start transforming your data. [Book a Free Consultation](https://analytium.com/30-minute-free-consultation) ## **During The Call, You Can Expect: ** - A brief analysis of your current data challenges - Recommendations tailored to your business needs - An overview of how Analytium’s solutions can drive your success ## Related Articles ##### [![The Practical Evaluation of AI Agents in Marketing](https://analytium.com/hs-fs/hubfs/Blog%20Images/July%2024/Practical%20Evaluation%20of%20AI%20Agents%20in%20Marketing.jpg?width=520&height=294&name=Practical%20Evaluation%20of%20AI%20Agents%20in%20Marketing.jpg) Blogs • November 27, 2024 The Practical Evaluation of AI Agents in Marketing](https://analytium.com/resource-hub/the-practical-evaluation-of-ai-agents-in-marketing) ##### [![Agentic AI in Sales](https://analytium.com/hs-fs/hubfs/Blog%20Images/July%2024/Agentic%20AI%20in%20Sales.jpg?width=520&height=294&name=Agentic%20AI%20in%20Sales.jpg) Blogs • November 8, 2024 Agentic AI in Sales](https://analytium.com/resource-hub/agentic-ai-in-sales) ##### [![Tackling The Productivity Challenge: Agentic AI](https://analytium.com/hs-fs/hubfs/Blog%20Images/July%2024/Tackling%20The%20Productivity%20Challenge%20Agentic%20AI.jpg?width=520&height=294&name=Tackling%20The%20Productivity%20Challenge%20Agentic%20AI.jpg) Blogs • November 3, 2024 Tackling The Productivity Challenge: Agentic AI](https://analytium.com/resource-hub/tackling-the-productivity-challenge-agentic-ai) [![Analytium Logo](https://analytium.com/hubfs/Analytium%20Logos/Logo%20250%20x%2066%20(5).svg)](https://analytium.com)   **[+44 (0) 203 574 8313](tel:+442035748313)**    [**contact@analytium.com**](mailto:contact@analytium.com)   4th floor, 86 - 90 Paul Street, London, United Kingdom, EC2A 4NE   ©2026 Analytium Ltd is a limited company registered in England and Wales. Company Registration No. 10214873 VAT: GB248508389  **[Privacy Policy](https://analytium.com/en/privacy-policy)**         - - - ```json { "@context" : "https://schema.org", "@type" : "BlogPosting", "author" : { "@type" : "Person", "name" : "Tim Matthews", "url" : "https://analytium.com/resource-hub/author/tim-matthews" }, "dateModified" : "2024-12-28T13:31:44.726Z", "datePublished" : "2024-09-04T19:59:19.000Z", "headline" : "Exploring Agents & AutoGen: Enhancing Workflow Automation", "image" : [ "https://analytium.com/hubfs/Blog%20Images/July%2024/Exploring%20Agents%20&%20AutoGen%20Enhancing%20Workflow%20Automation%20.jpg" ], "mainEntityOfPage" : { "@id" : "https://analytium.com/resource-hub/exploring-agents-and-autogen-enhancing-workflow-automation", "@type" : "WebPage" }, "publisher" : { "@type" : "Organization", "logo" : { "@type" : "ImageObject", "url" : "https://analytium.com/hubfs/logo-1.svg" }, "name" : "Analytium Ltd" } } ```