Martin Leitner

The value of EA is rarely in the Architecture Review Board. And none of your stakeholders will ever use the term “alignment.” Both standards and alignment are your job. Not theirs. When you are “doing” Enterprise Architecture, don’t explain your activities. Explain the results. Explain what you *hope* they will do with the results. Work with them to figure out how to fit *their* response to your results into *their* work processes. If they cannot use your results, either you are talking to the wrong person, or you are producing something useless. Architect, be useful.

129

41 Kommentare

If you are a M365 customer in Germany, Austria, Switzerland or anywhere in Europe, this update is for you. Please read it. Microsoft has just introduced #Anthropic models in #Copilot, and there are some important details you need to know before you enable them 🚨 Anthropic’s Claude Opus 4.1 is now available alongside OpenAI, but only in the Researcher agent and in Copilot Studio additional to Claude Sonnet 4 today. Great optionality, limited day to day impact for now. Microsoft announced in a side note, that this model availability of #Anthropic will come to M365 Copilot Chat experiences too! My clear CTA today: Switch it off 🔴 (or today: do not opt-in!) Why it matters Switching to Anthropic routes your prompts and responses OUTSIDE YOUR Microsoft’s managed environment. That means several enterprise protections you expect with Copilot do not apply. To be clear: no Microsoft Product Terms and DPA no Data residency and EU Data Boundary commitments no Microsoft audit and compliance controls and SLAs no Microsoft Customer Copyright Commitment My take for production tenants: Do not enable this yet. Block Anthropic at tenant level until protections reach parity. If you want to explore, do it only in a non production sandbox with non sensitive data and tight admin controls. By contrast, Microsoft‑hosted Microsoft 365 Copilot continues to operate under Microsoft’s privacy, security, and compliance commitments (including EU Data Boundary support). That’s the baseline many organisations chose Copilot for! Even in Copilot Studio, Anthropic models are treated as external and are not for production deployments How to block in the admin center Admin center > Copilot > Settings > Data access > AI providers for other large language models > Anthropic > Block provider My Bottom line Model choice is the right direction. Today, it is a promising preview for risk-aware companies, not a green light for regulated or risk averse organisations, especially in the EU. How are you approaching this change: pilot now or wait for protection parity? Raphael Köllner

175

39 Kommentare

“We switched to Event-Driven Architecture... and now we can’t debug anything.” That quote from a blog post might sound familiar if you’ve ever tried to replace everything with events. In this video, I break down exactly why Event-Driven Architecture can feel broken — and what actually causes the pain. https://lnkd.in/gaPxtUPV

57

2 Kommentare