As artificial intelligence technology progresses, creating agentic large language model (LLM) workflows is emerging as a key area of focus for developers looking to enhance their applications, with 92% of developers reporting that AI agents will advance their careers and 96% expecting these tools to improve the developer experience. These workflows empower LLMs to operate independently and make smart choices that enhance functionality across multiple applications, driven by low-code/no-code tools already used by 85% of developers to build and deploy AI agents efficiently. By leveraging innovative techniques like task decomposition, dynamic tool selection, and self-correcting architectures, developers can create LLM workflows that adapt to user requirements and contextual variations, with over 80% of developers believing AI agents will become as essential as traditional software tools.
At the Intel® AI DevSummit 2025, Microsoft Tech Lead Daron Yöndem presented a talk on "Building Agentic LLM Workflows with AutoGen," focusing on the importance of agentic systems in enhancing the performance of LLMs.
In this session, Daron explained that agentic systems enable multiple LLMs to collaborate, improving the quality of results through autonomous decision-making, collaboration, and adaptability. He highlighted the use of AutoGen, an open-source multi-agent framework, which facilitates the orchestration of agents and human oversight, ensuring efficient and high-quality outcomes.
Daron demonstrated various design patterns, including reflection, tool use, planning, and multi-agent collaboration, showcasing how these patterns can be implemented to achieve complex tasks. He provided practical examples and code demonstrations, emphasizing the flexibility and power of AutoGen in creating sophisticated workflows that leverage multiple agents and tools to solve intricate problems. Daron wrapped up the session with insights into real-world applications and challenges, underscoring the importance of strong guardrails and careful prompt engineering to ensure effective and reliable multi-agent systems.
"To harness the full potential of agentic AI,” said Daron, “we must prioritize strong guardrails and meticulous prompt engineering, ensuring that our multi-agent systems are both effective and reliable."
Watch the full video recording here to get more information on the practical implementation of agentic systems.
About the Speaker:
Daron Yöndem, with over 20 years in the tech industry, currently serves as a tech lead at Microsoft, focusing on Azure AI applications across 109 countries spanning Central and Eastern Europe, Middle East and Africa. His expertise spans solution architecture, SaaS development, and cloud native architecture. Before joining Microsoft, he founded a company and served as CTO for two startups, specializing in building SaaS products. He is also an author, with publications including a book on NoSQL databases. Connect with Daron on LinkedIn and X.
Hungry for More AI Knowledge?
Dive into more AI sessions from the Intel® AI DevSummit 2025 to learn from experts, explore the latest advancements and best practices, and take your projects to the next level.
We encourage you to also check out and incorporate Intel’s other AI/ML Framework optimizations and tools into your AI workflow and learn about the unified, open, standards-based oneAPI programming model that forms the foundation of Intel’s AI Software Portfolio to help you prepare, build, deploy, and scale your AI solutions.
Product Marketing Engineer bringing cutting edge AI/ML solutions and tools from Intel to developers.