AI-Generated Architectural Diagrams: Current Capabilities

Greg and Jeffrey discuss limits of diagram generation; simple wire diagrams work but detailed diagrams still fail.

Transcript

chevron_right

Play full episode

chevron_right

Transcript

Episode notes

Greg Leonardo is a Principal Cloud Architect, Certified Azure Solution Architect Expert, and Microsoft Azure MVP with over two decades of experience driving high-impact cloud transformations for global enterprises. Greg’s work sits at the crossroads of architecture, innovation, and mentorship, where he crafts scalable, cost-efficient, and future-ready solutions that align technology with business strategy. He’s a recognized authority in Azure-native architecture, FinOps optimization, and AI integration, known for delivering measurable results—from reducing cloud spend to accelerating DevOps agility through Terraform and infrastructure as code. His leadership extends beyond technical execution, as he actively cultivates cloud talent and fosters a culture of continuous improvement. A former Gulf War veteran, Greg brings discipline, resilience, and strategic foresight to every initiative — whether building global cloud ecosystems or leading high-performance engineering teams. He’s also a passionate advocate for the cloud community, sharing insights daily through technical writing and open mentorship.

Topics of Discussion:

[5:46] Greg discusses his approach to AI, emphasizing responsible AI and the importance of understanding its limitations.

[8:12] AI’s journey through the Hype Cycle.

[11:09] Greg introduces the concept of AI as “virtual employees.”

[16:44] Greg talks about how he uses both Microsoft 365 Copilot and ChatGPT Enterprise to compare their capabilities.

[18:36] How Greg tests and compares AI tools.

[20:21] Comparing AI to previous search and research methods.

[23:31] Improving AI results through better prompts.

[26:17] The agent framework, sharing, and limitations of AI agents.

[36:47] AI’s ability to generate architectural diagrams.

[41:49] Exploring autonomous coding agents.

[41:45] Why we aren’t quite ready for the “fire and forget” side.