Architectural Alchemy - Transforming Systems through AI-Driven Design

Breakout session presented at KPMG Technology Days 2025 in Zurich, Switzerland.

The presentation breakdown, including additional resources, is available at https://blog.marceloancelmo.com/architectural-alchemy

AI adoption is exploding, and without a solid architectural foundation it leads to chaos. How can enterprises scale AI adoption responsibly?

In this presentation, we explore the case for treating AI as a first-class architectural element. It brings together Artificial Intelligence, software architecture, and DevOps practices to show how intelligence-driven systems change not only how software is built and delivered, but how organizations make decisions.

You will leave with three takeaways: establish an AI security and governance framework, treat explainability as a key quality attribute, and remember that your AI is only as good as your data.

By embracing AI responsibly, organizations can unlock unprecedented levels of efficiency, scalability, and innovation, all while managing risks and ethical concerns. This session provides actionable strategies to integrate AI into systems and leverage DevOps practices, helping organizations stay agile and competitive in an era where intelligence-driven software is not optional but essential for driving meaningful business outcomes.

If you enjoyed these slides, please let me know at https://marceloancelmo.com.

Transcript

1. Architectural Alchemy Transforming Systems through AI-Driven Design Marcelo Sousa Ancelmo Head of Solution Architecture KPMG Technology Days 2025 KPMG Switzerland 27th of August 2025
2. The AI Transformation Journey forSystems Image Source: claude.ai  © 2025 KPMG AG, a Swiss corporation, is a subsidiary of KPMG Holding AG, which is a member of the KPMG global organization of independent firms affiliated with KPMG International Limited, a private English company limited by guarantee. All rights reserved. Document Classification: KPMG Public 2
3. Architectural Priorities Foundational Considerations • Why is more important than How • Data and Content • Privacy by Design • Modularity • Model Lifecycle Management • Model Registry Good Practices • Separation of training and inference infrastructure • Embed AI into Observability • AI-as-a-Service Source: Lifecycle and Key Dimensions of an AI System. Modified from OECD (2022) - NIST - Artificial Intelligence Risk Management Framework (AI RMF 1.0) • Human-in-the-Loop for Supervision © 2025 KPMG AG, a Swiss corporation, is a subsidiary of KPMG Holding AG, which is a member of the KPMG global organization of independent firms affiliated with KPMG International Limited, a private English company limited by guarantee. All rights reserved. Document Classification: KPMG Public 3
4. Data is at the Center Quality Privacy Data Labelling Classification © 2025 KPMG AG, a Swiss corporation, is a subsidiary of KPMG Holding AG, which is a member of the KPMG global organization of independent firms affiliated with KPMG International Limited, a private English company limited by guarantee. All rights reserved. Document Classification: KPMG Public 4
5. Intelligent Content Management Content Classification •Auto-tagging •Sentiment Analysis Content Generation •Content Creation Summarization •Insights Extraction Search •Embeddings •Combine LLMs with traditional search Compliance •Policy enforcement Personalization •Context aware delivery © 2025 KPMG AG, a Swiss corporation, is a subsidiary of KPMG Holding AG, which is a member of the KPMG global organization of independent firms affiliated with KPMG International Limited, a private English company limited by guarantee. All rights reserved. Document Classification: KPMG Public 5
6. Explainability as a Quality Attribute Transparency Interpretability Reasoning Provenance Accountability © 2025 KPMG AG, a Swiss corporation, is a subsidiary of KPMG Holding AG, which is a member of the KPMG global organization of independent firms affiliated with KPMG International Limited, a private English company limited by guarantee. All rights reserved. Reproducibility Document Classification: KPMG Public 6
7. Security – The Lethal Trifecta The lethal trifecta of capabilities is: •Access to your private data—one of the most common purposes of tools in the first place! •Exposure to untrusted content—any mechanism by which text (or images) controlled by a malicious attacker could become available to your LLM •The ability to externally communicate in a way that could be used to steal your data (I often call this “exfiltration” but I’m not confident that term is widely understood.) Source: The lethal trifecta for AI agents: private data, untrusted content, and external communication And thereismore: • Prompt Injection • Data Poisoning • Model Extraction •… © 2025 KPMG AG, a Swiss corporation, is a subsidiary of KPMG Holding AG, which is a member of the KPMG global organization of independent firms affiliated with KPMG International Limited, a private English company limited by guarantee. All rights reserved. Document Classification: KPMG Public 7
8. Governance, Risk & Compliance (GRC) Source: NIST - Artificial Intelligence Risk Management Framework (AI RMF 1.0) © 2025 KPMG AG, a Swiss corporation, is a subsidiary of KPMG Holding AG, which is a member of the KPMG global organization of independent firms affiliated with KPMG International Limited, a private English company limited by guarantee. All rights reserved. Document Classification: KPMG Public 8
9. Moving Towards a Responsible and Ethical AI Image Source: Google Gemini  © 2025 KPMG AG, a Swiss corporation, is a subsidiary of KPMG Holding AG, which is a member of the KPMG global organization of independent firms affiliated with KPMG International Limited, a private English company limited by guarantee. All rights reserved. Document Classification: KPMG Public 9
10. AI-Driven Design is Easy – Pick your Poison Image Source: What is Model Context Protocol (MCP)? And why is everybody talking about it? Image Source: https://sierra.ai/blog/the-challenge-with-rolling-your-own-agent © 2025 KPMG AG, a Swiss corporation, is a subsidiary of KPMG Holding AG, which is a member of the KPMG global organization of independent firms affiliated with KPMG International Limited, a private English company limited by guarantee. All rights reserved. Document Classification: KPMG Public 10
11. AI Powered Takeaways Establish a solid AI Security and Governance Framework Explainability is a key quality attribute for AI Systems Your AI is as good as your Data © 2025 KPMG AG, a Swiss corporation, is a subsidiary of KPMG Holding AG, which is a member of the KPMG global organization of independent firms affiliated with KPMG International Limited, a private English company limited by guarantee. All rights reserved. Document Classification: KPMG Public 11
12. “You don’t have to be an AI expert, but you must be an AI explorer.” Andrew Ng @marceloancelmo #daretodo #together4better #AI+thekitchensink © 2025 KPMG AG, a Swiss corporation, is a subsidiary of KPMG Holding AG, which is a member of the KPMG global organization of independent firms affiliated with KPMG International Limited, a private English company limited by guarantee. All rights reserved. Document Classification: KPMG Public Photo by Janko Ferlič on Unsplash 12