Vesa Juvonen

There are very few real Enterprise Architects today. That’s not a criticism. It’s an architectural reality we quietly inherited. Somewhere along the way, Enterprise Architecture stopped designing the enterprise… and started managing applications. Instead of defining capability structures, behavioral boundaries, and integration intent, EA gradually became responsible for: • application portfolio alignment • integration governance • standards enforcement • landscape documentation • constraint negotiation Architecture became reactive. Applications became structural. And enterprises became application-defined instead of architecture-defined. Over time, applications created invisible “hard walls”: • logic boundaries • data ownership boundaries • behavioral silos • integration friction layers Rather than redesigning these boundaries, we built compensating mechanisms around them: integration hubs canonical data models middleware layers governance overlays These help manage fragmentation — but they don’t eliminate it. Then Data Architecture stepped in to restore coherence. It introduced: • semantic layers • master data management • canonical enterprise models But data defines meaning, not behavior. Applications still control behavior. And Enterprise Architecture ended up sitting between the two — governing structure without truly shaping it. That’s the gap. True Enterprise Architecture was never meant to catalogue systems. It was meant to define enterprise behavior. A real Enterprise Architect designs: • capability structures • enterprise semantics • integration strategy • decision authority boundaries • operating model logic Applications should implement architecture. Not define it. Today, something important is changing. With agentic platforms, intent-driven automation, and closed-loop architectures, Enterprise Architecture is becoming executable again. Instead of: Architecture → documentation We now have: Architecture → policies → agents → automation → validation → learning loops This is where Enterprise Architecture returns to its original purpose: Designing how the enterprise behaves. Not describing what systems exist. The future Enterprise Architect will not just align systems. They will orchestrate enterprise behavior. And that’s a very different role.

9

1 kommentti

👋 Hi everyone, We’ve updated the Greengage DB roadmap and outlined the main directions for the product’s development in the upcoming phases. Keep in mind that the roadmap doesn’t include everything we’re currently working on - some features still require additional research and will be added later. You can always check out the latest version of the Greengage DB roadmap on the project website - https://lnkd.in/ehaCaYHt . 🚀 P.S. We’re also planning to ship some significant updates this week - stay tuned for more details - https://lnkd.in/eA7WcYBU. Greengage DB Team - https://lnkd.in/eJhv5Uyp #GreengageDB #OpenSource #PostgreSQL #Greenplum #SoftwareEngineering #Database #BigData #ProductRoadmap

20

M365 Copilot doesn't have a usage problem. It has a default problem. Most orgs assign Copilot licenses the same way they assign Exchange mailboxes - to every user at onboarding, no review, no expiry. The assumption is that someone will use it eventually. I saw this pattern at scale before stepping back from work. Blanket M365 Copilot assignments across an entire org. One IT admin managing it. No audit loop. I built a PowerShell script to pull provisioning data and surface idle seats. The numbers weren't rounding errors. They were a real budget conversation - because idle didn't mean low activity. It meant zero. The root issue isn't that Copilot is overpriced. It's that license assignment is a one-click action and license review is a manual project nobody schedules. Microsoft's defaults push toward coverage, not fit. Automating the audit loop changes the incentive. When leadership sees a weekly report showing idle seats with a cost figure attached, the conversation shifts from 'we need more AI tools' to 'we need to actually use what we have.' If you've run a Copilot or M365 license audit: what was the idle-seat rate, and what finally got leadership to act on it? #M365Admin #CopilotForMicrosoft365 #PowerShell #LicenseManagement #FinOps

1

The EASIEST way to spot a Junior developer? Look at their pubspec.yaml file. If you have 45 different packages installed for basic things like formatting dates or adding a shadow... your app is a TICKING TIME BOMB. Every time you add a package, you are marrying that author's code. ➔ When they stop updating it, your app BREAKS. ➔ When version conflicts happen, you spend hours in DEPENDENCY HELL. Senior developers understand that the cost of MAINTAINING a package is often higher than the cost of just writing the logic yourself. If a feature takes 20 lines of pure Dart code, WRITE THE CODE. Stop outsourcing your core logic to random internet strangers. Developers: What is one package you used to rely on, but now you just write from scratch? #FlutterDev #Dart #SoftwareArchitecture #CleanCode #MobileDevelopment

10

2 kommenttia

Ghost Architect™ v7.0.0 is here. This release belongs to the 3,933 people already running Ghost Architect™. The feedback you sent in shaped what shipped. New features, expanded capabilities, lower API costs, and better reporting are all in this version because users asked for them. To mark the release, the first 100 new subscribers and current subscribers upgrading get 50% off the first month. Code LAUNCH50 at checkout. A word on security and privacy before anything else, because it is the question every serious developer or agency asks first. Ghost Architect™ runs on your local machine. It does not download your code. It does not save your code. It does not transmit your code to any Ghost Architect™ server. Your client's code stays yours. The headline addition is Ghost Portal, available on Ghost Team and Ghost Enterprise. Portal is a browser-accessible view into the data and performance of every project, every team, every effort. Open it from anywhere. At home on a Sunday. In a board meeting when someone asks a question and you need the answer now, not after you dig through Jira tickets. When it is time to move on findings, Portal exports directly to the project management software your team already uses through CSV. Portal stores findings metadata only, never source code. That is what changes for the people leading the work. Here is what changes for the people running it. Ghost Open, the free tier that gets you started, is now structured the way the work actually happens. Question mode is unlimited. Ask your codebase anything, as often as you need. Recon mode is unlimited. Get a real engagement plan on every prospect, every project, every codebase you are sizing up. And you now have four free deep reports to spend across the modes that drive real decisions: Points of Interest, Blast Radius, Conflict Detection, and Prompt Triage. Use them however you want. Four reports is enough to put Ghost Architect™ to work on the codebase that is on your screen right now, not just kick the tires. That is the point. When the reports prove their value, the upgrade path is built for the way you actually work. Ghost Pro is a single seat, built for the architect, consultant, or fractional CTO who operates solo. Ghost Team is five seats, for the small team running shared engagements. Ghost Portal is included. Ghost Enterprise scales beyond that. Multiple groups, multiple projects, multiple countries, one organization, one source of truth. Ghost Portal is included. If you have been running Ghost Architect™, this is the version worth updating to. If you have not yet, the install command and pricing are in the first comment. Ernst

1

1 kommentti

How to use a WebView2 control in a VC++ MFC dialog application I could not find a single example of how to implement a WebView2 in MFC. Even the official Microsoft Webview2 examples have it for WPF, Win32, UWP etc. but not for MFC. So I looked up the current solutions and customized them for use in MFC. Source available on Github: https://lnkd.in/gc75EBhr

1

AI agents are rapidly automating the implementation phase. So, where does the value shift? 🧐 The teams that win in the future will double down on User Experience—both Pre and Post. Pre-Execution (The Intent): Deeply understanding the user problem. Designing with clarity. Shaping the work so agents have the right context to build what actually matters. Post-Execution (The Outcome): Obsessing over the final quality. Reviewing, testing, and ensuring the solution actually solves the human problem it was meant to address. When coding becomes a commodity, clarity becomes the craft. The best teams won't just be the ones who code the fastest; they will be the ones who measure and deliver —pre and post—that the user's needs are actually met.

3

Is it just me, or are a lot of “new” Microsoft enterprise tools starting to feel more packaged than truly new?

4

Why (pure) functional programming matters by Christoffer Ekeroth It is well-known that as a program grows, it becomes more and more difficult to understand and extend—it Is there anything we can do to fix it?In this talk you will learn how seemingly innocuous concepts like nullability, side effects and exceptions make your code base harder to understand. You will also learn how functional programming techniques can help you write programs that are easy to reason about, no matter their size. https://lnkd.in/dBKrSdxY #funcprogsweden #functionalprogramming

58

FlowCentric Processware seamlessly provides insight in average process execution time with exception reporting such as process bottlenecks, or number of reworks per function, process tasks, or user. This enables organisations to optimise processes and eliminate inefficiencies where applicable.

6

Three weeks ago I shared the architecture. 3 days ago I talked about the rebuild. Now I stress-tested it against real data. And I found exactly what I expected to find — silent failures that would have caused serious problems in production. Here's what was broken: A cross-tenant data leak. One organization's vendor GST was matching another organization's tax engine query. In a multi-client AP system, that's not a bug. That's a liability. Non-atomic journal entries. The debit line and credit line were being written as separate database queries. A crash between them would leave the books inconsistent. No error. No alert. Just wrong numbers. Self-reported reviewer roles. Anyone could send a POST request claiming CFO authority and approve a ₹5 lakh invoice. The approval tier logic was correct — but it was trusting the request body to tell it who was asking. LLM normalization gap. The same vendor submitting as "Acme Corp", "ACME CORP." and "Acme Corporation" was being treated as three different vendors by the duplicate check. Real invoices were slipping through as new submissions. A dead code fallback. The subtotal + tax calculation that was supposed to reconstruct a missing total amount? It was written after the null guard that returned early. It never executed. Not once. None of these failed loudly. That's what makes them dangerous. After fixing the foundation I added a fraud scoring engine — six signals that run on every invoice before it reaches compliance. Round number amounts. First-time vendors above threshold. Backdated invoices. PO number reuse across multiple invoices. Rapid resubmission from the same vendor. Amount spikes above historical average. Each signal contributes to a risk score. Below 30 — the invoice moves forward automatically. Between 30 and 60 — it waits for human review, but escalates itself after 4 hours. Above 60 — auto-blocked. The system now resolves most exceptions without human intervention. When it can't, it knows exactly how long to wait before escalating. Next — LangGraph and OpenAI migration. The state machine stays. The intelligence gets upgraded. #SystemDesign #Fintech #BackendDevelopment #CyberSecurity #APAutomation #BuildingInPublic

12

1 kommentti

Umbraco was sowing the seeds of love at #Codegarden2025 – and it was music to everyone’s ears. The annual conference for the #opensource .NET community just played to a packed house in Odense, Denmark – celebrating its 20th year with a setlist of amazing speakers, sessions, workshops, and announcements. Attendees were on site and dialed in virtually, but there were more than developers in the crowd. Codegarden offered a rich slate of relevant learning and #content for designers, strategists, and partners, and we heard from leaders and luminaries about a range of topics. There was a lot to cover, from key feature enhancements to the anointing of this year’s crop of Umbraco MVPs. But a few things stole the spotlight, and I connected with Umbraco's CTO, Filip Bech-Larsen, and Senior Product Manager, Bolette Kern, for more insight. At the top of the list was #AI and some news around Umbraco’s marketplace roadmap and the introduction of an #MCPserver (something capturing the #DXP zeitgeist in the #agenticAI era). I also got an up-close look at the new Umbraco Compose, which provides a new layer of composable orchestration – and some exciting possibilities for builders. In my overview on CMS Critic, I explore these topics with Filip and Bolette. But the technical pieces are also balanced by this community's culture, which continues to write its own soundtrack by cultivating a “Friendly CMS” experience and focusing on critical issues like sustainability. “I love to see the difference we can make in the world,” Filip says. That speaks volumes about the kind of company this is – and why its community is still head over heels about it. Read it here: https://lnkd.in/e39cDm3j

32

7 kommenttia

Having a ton of static diagrams of you Enterprise Architecture and challenged to keep them up to date. Ardoq has just released a new way to fast bring those documents alive. Check out Christoffer Rou Nowicki short demo video below. Really cool! 🚀

6

I often think these days about horrible, disfunctional and nightmarish codebases I had to maintain in my past. And then I look at how fast and efficient AI tools generate code at our command. And then I smile. Bitterly. My job is secured. Not sure about yours tho.

11

📢 I am looking forward to the upcoming NNUG OSLO meetup on Monday June 2nd with Tore Nestenius as our speaker! 🎙️ His talk is named Top Authentication Fails in ASP.NET Core (and How to Avoid Them) 🚨 In this talk, Tore will share the most common authentication mistakes developers make in ASP.NET Core, especially around OpenID Connect. These include misconfigurations, protocol misunderstandings, and insecure practices. The session highlights anti-patterns, explains how they can lead to security vulnerabilities, and offers best practices to help developers build more secure and reliable authentication systems. 📅 Monday, June 2nd 🕰️ 16:30 CEST 🏠 Rebel (meeting room Niels Henrik) 🥳 There is still time to sign-up on our meetup page to join this session https://lnkd.in/d4hPXQ9h #mvpbuzz #mctbuzz #azure #security

11

2 kommenttia