The Practical CTO

The EU AI Act is no longer a future concern. As of August 2026, high-risk AI system obligations are in full effect, and enterprises operating in or selling into the EU market are either compliant, in remediation, or quietly hoping nobody notices. The last category is shrinking fast as national supervisory authorities begin their first formal enforcement cycles. I've spent the past several months helping organizations work through their AI Act readiness assessments, and the gap between what leadership thinks their compliance posture is and what it actually is tends to be significant. This post is about closing that gap practically — not rehashing the regulation's structure, which you can read in the official text. Why Most Enterprise AI Act Assessments Are Incomplete The first problem is scope. Most organizations have catalogued their "AI systems" based on what the IT department built or purchased as AI. That misses a large category: ML models embedded in existing...

Most AI copilots arrive with a press release and a demo. SAP Joule arrived with something rarer: actual enterprise context. Having watched it roll out across several large S/4HANA deployments over the past year, I can say the gap between the marketing narrative and what practitioners are discovering on the ground is worth examining carefully. What Joule Actually Is (And What It Isn't) SAP positions Joule as a "generative AI copilot embedded across the SAP ecosystem." That's accurate but incomplete. What makes Joule different from a generic LLM wrapper isn't the model underneath — it's that it has read access to your SAP data graph and understands business object semantics. When you ask Joule "why did this purchase order get blocked?", it doesn't search the internet. It queries your actual procurement data. The limitation worth stating upfront: Joule is only as good as your data quality and your SAP configuration. In organizations where master...

Why Joule in 2026 Is Different From the Version Announced in 2023 I want to be direct about something upfront: SAP Joule was announced with considerable fanfare in September 2023, and the gap between that announcement and the production capabilities available at launch was wide enough to generate genuine skepticism among enterprise customers. I shared some of that skepticism. Early Joule was a promising research demonstration; it was not an enterprise production tool. By mid-2026, that has changed materially. Joule is now embedded across the SAP product portfolio in ways that are genuinely useful to the finance, HR, supply chain, and IT teams I work with — not just as a conversational novelty but as a system that reduces measurable friction in daily work. The change hasn't been a single dramatic release; it's been a steady expansion of the contexts in which Joule can actually access live enterprise data, take actions, and generate outputs that save real time. This p...

Most internal developer platforms fail. Not because the engineers who built them were incompetent, not because the underlying technology was wrong, and not because the organization lacked budget. They fail because the teams that built them treated platform engineering as a purely technical problem — and it is not. In 2026, the conversation around Internal Developer Platforms (IDPs) has matured considerably. We have more open-source tooling than ever before, better reference architectures, and a growing body of case studies. And yet, adoption rates remain stubbornly low at many companies. Surveys consistently show that a significant portion of engineering organizations have some form of an IDP initiative, but fewer than half of developers actually use it as their primary workflow. I have spent the last several years working on and with platform teams — observing what gets traction, what gets ignored, and what gets actively resisted. The patterns are consistent enough that I am ...