For decades, we have governed technology as infrastructure—managing it through security protocols, uptime, and access controls. But as we enter the era of relational AI, this paradigm is beginning to fail.

This podcast explores the groundbreaking case for Artificial Human Resources (AHR), a new governance framework for intelligent systems designed with empathy-integrated architecture. Drawing from the latest 2026 working paper, we discuss why treating a sophisticated AI agent as a mere "tool" is no longer operationally sufficient when that agent makes decisions affecting human dignity.

In this series, we break down:

• The Empathy Threshold: Why systems that model the "whole person"—their work, health, and family—require oversight analogous to human resources management.
• The Governance Gap: Why current enterprise standards like TOGAF and IAM are architecturally incomplete for governing agents that learn and adapt over time.
• The AHR Lifecycle: A deep dive into the operational stages of AHR, from ethical onboarding and relational performance evaluation to the responsible retirement of agents humans have grown to trust.
• A New Organizational Chart: How the "Agentic Enterprise" must integrate HR specialists, psychologists, and ethicists into the core of technical systems design.

As we externalize intelligence into machines, the qualities that remain distinctively human—empathy, moral judgment, and relational wisdom—become our most valuable assets. Join us as we explore how AHR ensures that the power of AI becomes constructive rather than destructive, forcing us to mature philosophically as much as we have technologically.

We are currently living through the most consequential window in human history—a period where AI is powerful enough to reshape our world, but still within our window of control. In this podcast, we explore the transition from AI as a tool to AI as a participatory member of human collectives.

Drawing on research from Stanford HAI, McKinsey, and the World Economic Forum, we break down the rise of Sovereign AI, where nations like the U.S., Saudi Arabia, and India are investing hundreds of billions to ensure their cultural and legal values are embedded in the "civilizational infrastructure" of the future.

We also tackle the "deepest fault line" in AI development: the consciousness question. As companies like Anthropic begin hiring AI welfare researchers and legal scholars argue for future AI personhood, we examine a world where humans may eventually move from controlling AI to negotiating with it.

Join us as we map out the next fifty years of human-AI coevolution, from the formation of global governance blocs to the emergence of deeply entangled, semi-autonomous regional collectives. The decisions we make today about audit tools and training methodologies are not just technical—they are civilizational.

This episode explores the architectural impact of SAP’s latest Sapphire announcements and the broader shift toward AI-enabled enterprise systems.

It examines how enterprise AI is moving toward better grounding in business data using knowledge graphs and retrieval-augmented generation (RAG), helping large language models operate with more accurate and contextual understanding of enterprise information.

The discussion also covers the challenge of unifying structured ERP data with unstructured enterprise knowledge from tools like Slack and Microsoft Teams, and the difficulty of turning informal work patterns into governed, usable business intelligence.

It looks at the rise of conversational AI interfaces such as SAP Joule, the emergence of agent-based workflows across enterprise systems, and the importance of maintaining auditability, security, and compliance in automated decision-making.

Finally, it highlights why modern data architectures and open table formats such as Apache Iceberg are increasingly important for enterprise AI readiness, alongside the ongoing challenge of modernizing legacy ERP landscapes.

Generative AI agents mark a significant leap forward from traditional language models, offering a dynamic approach to problem-solving, and the future of AI is considered agentic. This podcast serves as a "102" guide for developers seeking to transition their AI agent proofs-of-concept into reliable, high-quality production systems.

We delve into the crucial practices of Agent and Operations (AgentOps), a subcategory of GenAIOps that focuses on the efficient operationalization of agents. AgentOps incorporates DevOps and MLOps principles while adding agent-specific components like tool management, orchestration, memory, and task decomposition. We emphasize that metrics are critical; successful deployment requires tracking not just business KPIs (like goal completion rate) but also detailed application telemetry and human feedback.

A core focus is Agent Evaluation, which is essential for bridging the gap to production-ready AI. We explore the three key components of evaluation:

1. Assessing Agent Capabilities against public benchmarks to identify core strengths and limitations.
2. Evaluating Trajectory and Tool Use by analyzing the steps an agent takes toward a solution using ground-truth metrics like Exact Match, Precision, and Recall.
3. Evaluating the Final Response using custom success criteria and autoraters (LLMs acting as judges).We also stress the necessity of Human-in-the-Loop evaluation to assess subjective qualities like creativity and nuance, and to calibrate automated evaluation methods.

Furthermore, we explore advanced systems, starting with Multi-Agent Architectures, where multiple specialized agents collaborate to achieve complex objectives. These architectures offer enhanced accuracy, efficiency, scalability, and better handling of complex tasks. Key multi-agent design patterns are discussed, including the Hierarchical Pattern (a manager coordinating workers), the Diamond Pattern (responses moderated before output), Peer-to-Peer (agents hand off queries to one another), and the Collaborative Pattern (multiple agents contributing complementary information). We use Automotive AI as a compelling case study to illustrate these real-world multi-agent implementations.

We examine Agentic RAG (Retrieval-Augmented Generation), a critical evolution that uses autonomous agents to iteratively refine searches, select sources, and validate information, leading to improved accuracy and context-aware responses. Importantly, we cover the need to optimize underlying search performance (e.g., semantic chunking, metadata enrichment) before complex RAG implementation.

Finally, we discuss the role of agents in the enterprise, where knowledge workers become managers of agents who orchestrate automation and assistant agents. We detail enterprise platforms like Google Agentspace and propose the evolution toward 'Contract adhering agents,' which standardize tasks with clear deliverables, validation mechanisms, negotiation, and subcontracts for high-stakes problem-solving. Tune in to understand the tools and techniques—including Vertex AI Agent Builder, Eval Service, and the Gemini models—to confidently build, evaluate, and deploy the next generation of intelligent applications.

Are your AI initiatives stalling at proof-of-concept? Up to 95% of AI pilots fail to deliver measurable profit impact, often due to fragmentation, lack of governance, and absence of strategic alignment. In this episode, we explore how Enterprise Architecture (EA) becomes the essential backbone for turning isolated AI experiments into scalable, sustainable business capabilities.

Join us as we dive into how EA acts as a dynamic capability that enables organizations to sense, seize, and transform around GenAI—all while forging real business value. We unpack the Architecture of AI Transformation framework, highlighting how to move beyond incremental automation toward a new frontier of Collaborative Intelligence, where human judgment and AI scale merge effectively.

You’ll learn how EA operationalizes scalable AI across four pillars: Business Alignment, System Integration, Process Awareness, and Governance & Accountability. We’ll also unpack how composable AI architectures and “glass-box” governance prepare you for regulatory demands like the upcoming EU AI Act (August 2026).

If you’re responsible for AI strategy, digital transformation, or enterprise architecture, this episode gives you practical insights and research-based models to embed AI not just as an experiment—but as a core, governed, and high-impact capability.

What you’ll get:

• Why AI pilots so often fail to scale — and how EA solves that gap.

• How to treat EA as a dynamic capability: sensing opportunities, seizing demand, transforming operations.

• How to think beyond process automation toward collaborative intelligence (human + machine).

• The four pillars of scalable, governed AI: business alignment, system integration, process awareness, and governance.

• Real-world implications for reuse, composable architecture, workflow redesign and regulatory readiness (EU AI Act).

Credits & References:
Based on foundational research from:

• Ettinger, A. (2025). Enterprise Architecture as a Dynamic Capability for Scalable and Sustainable Generative AI Adoption. Warwick Business School.

• Wolfe, D. A., Choe, A., & Kidd, F. (2025). The Architecture of AI Transformation: Four Strategic Patterns and an Emerging Frontier.

• Zeman, B. (2025). Why Enterprise Architecture Is the Missing Link in Scalable AI. Built In.

• Built In. Enterprise Architecture for Scalable AI Implementation.

In this episode of State of AI, we explore how artificial intelligence is reshaping the foundations of the software industry. Drawing on insights from McKinsey’s 2025 report Upgrading Software Business Models to Thrive in the AI Era, we unpack how AI is transforming SaaS from a tool that enables work to a platform that performs it.

Learn why traditional subscription models are evolving toward consumption-based pricing, how companies like Salesforce, Adobe, and ServiceNow are experimenting with AI-driven monetization, and what it takes to align pricing, GTM strategy, and customer value in the new AI economy.

We’ll also dive into:

• The $4.4 trillion productivity potential of AI in enterprise software

• Real-world examples of AI monetization success and failure

• Why change management now costs 3x more than model development

• How CFOs and GTM teams can prepare for AI-driven revenue models

• The new role of AI agents in redefining customer value and pricing

If you’re a tech executive, SaaS founder, or digital strategist, this episode will help you understand how to adapt your business model to capture the next software supercycle.

Join us for an insightful exploration into the world of Reasoning LLMs, drawing on the expertise of Sebastian Raschka, PhD. This episode demystifies how Large Language Models (LLMs) are being refined to excel at complex tasks that require intermediate steps, such as solving puzzles, advanced mathematics, and challenging coding problems, moving beyond simple factual question-answering.

We'll uncover the four main approaches currently used to build and improve these specialised reasoning capabilities:

• Inference-time scaling: Discover how techniques like Chain-of-Thought (CoT) prompting encourage LLMs to generate intermediate reasoning steps, mimicking a 'thought process' and often leading to more accurate results on more complex problems. This approach increases computational resources during inference, making it more expensive.
• Pure Reinforcement Learning (RL): Learn about the surprising emergence of reasoning behaviour from pure reinforcement learning, as demonstrated by DeepSeek-R1-Zero. This model was trained exclusively with RL, without an initial supervised fine-tuning (SFT) stage, using accuracy and format rewards to develop basic reasoning skills.
• Supervised Fine-tuning (SFT) + Reinforcement Learning (RL): Understand this key approach for building high-performance reasoning models, exemplified by DeepSeek's flagship R1 model. This method refines models with additional SFT stages and further RL training, building upon "cold-started" pure RL models.
• Pure SFT and Distillation: Explore how smaller, more efficient reasoning models can be created by instruction fine-tuning them on high-quality SFT data generated by larger, stronger LLMs. This approach is particularly attractive for creating models that are cheaper to run and can operate on lower-end hardware.

We'll also discuss when to use reasoning models – they are ideal for complex challenges but can be inefficient, more verbose, and expensive for simpler tasks, sometimes even being "prone to errors due to 'overthinking'". The episode provides valuable insights from the DeepSeek R1 pipeline as a detailed case study and touches upon comparisons with models like OpenAI's o1. Plus, get tips for developing reasoning models on a limited budget, including the promise of distillation and innovative methods like 'journey learning', which includes incorrect solution paths to teach models from mistakes. Tune in to navigate the rapidly evolving landscape of reasoning LLMs!

Join us for an insightful exploration into the cutting-edge design of today's Large Language Models. Seven years on from the original GPT architecture, have we truly seen groundbreaking changes, or are we simply refining existing foundations? This podcast focuses on the architectural developments that define flagship open models in 2025, moving beyond benchmark performance or training algorithms.

In this episode, we'll unpack the key ingredients contributing to LLM performance, examining how developers are pushing the boundaries of efficiency, memory management, and training stability. Discover the evolution and intricacies of:

• Attention Mechanisms: From Multi-Head Attention (MHA) to the more efficient Grouped-Query Attention (GQA), and innovative approaches like Multi-Head Latent Attention (MLA) used in DeepSeek-V3, which compresses key and value tensors for memory savings. We also delve into Sliding Window Attention from Gemma 3, which restricts context size for local efficiency.
• Normalization Layers: Explore the shift from LayerNorm to RMSNorm and the crucial placement of these layers (Pre-Norm, Post-Norm) as seen in OLMo 2 and Gemma 3, including the addition of QK-Norm for enhanced training stability.
• Mixture-of-Experts (MoE): Understand why this approach has seen a significant resurgence in 2025. Learn how MoE, as implemented in models like DeepSeek-V3, Llama 4, and Qwen3's sparse variants, allows for massive total parameter counts (e.g., DeepSeek-V3's 671 billion parameters) while activating only a small subset (e.g., 37 billion) per inference step for remarkable efficiency.
• Positional Embeddings: Discover how positional information is handled, from rotational positional embeddings (RoPE) to the radical concept of No Positional Embeddings (NoPE) in SmolLM3, which aims for better length generalization.

We'll compare the structural nuances of leading models such as:

• DeepSeek-V3: A massive 671-billion-parameter model known for MLA and MoE with a shared expert.
• OLMo 2: Notable for its transparency and specific RMSNorm placements for training stability.
• Gemma 3 & 3n: Featuring sliding window attention for KV cache memory savings and unique normalization layer placements; Gemma 3n also introduces Per-Layer Embedding and MatFormer concepts.
• Mistral Small 3.1: Prioritizing lower inference latency through custom tokenizers and specific architectural choices.
• Llama 4: Adopting an MoE approach similar to DeepSeek-V3 but with its own distinct expert configuration.
• Qwen3: Available in both dense and MoE variants, offering flexibility for various use cases and moving away from shared experts in some MoE configurations.
• SmolLM3: A compact 3-billion-parameter model exploring the effectiveness of NoPE.
• Kimi K2: An impressive 1 trillion parameter model, building on the DeepSeek-V3 architecture with more experts and fewer MLA heads, setting new standards for open-weight performance.

Tune in to understand the intricate design decisions driving the next generation of large language models.