Broentech

01/06/2026

Most founders treat client work as a distraction from building product. We've spent sixteen years proving the opposite.

Every consulting engagement is a stress test. Real systems break in ways your roadmap never predicted. And the ideas that don't survive contact with production? Those are the ones you want killed early.

Our flywheel is simple: client problems sharpen product ideas. Product development pushes what we bring back to clients. One loop feeds the other 🔄

Swipe through to see how this filter actually works 👇

29/05/2026

Most AI features feel bolted on because they were.

After 16 years of shipping production AI systems, we kept seeing the same pattern. Teams build the product first, then try to squeeze AI in at the end. It rarely works.

So we built a framework around what actually does. We call it the Core-First Method, and it comes down to three steps:

1. Architect the AI core before writing any product code
2. Build every interface and workflow around the model
3. Ship to production with AI running from day one

The difference is simple. When AI is the product spec, not a feature request, everything downstream gets easier. Data flows make sense. Interfaces serve the model. Deployment isn't a surprise.

Swipe through for the full breakdown 👇

27/05/2026

"Just add an AI layer on top."

Every product roadmap we review in 2026 starts with that sentence. After sixteen years of shipping production systems, we can tell you exactly where that road ends.

About four months in. That's when the data flow breaks, the UX starts fighting the architecture, and that AI feature that looked great in a demo becomes the thing your team spends 60% of their time patching.

We see it again and again. A team bolts a model onto an existing product. Early results look promising. But the architecture underneath was never designed to let AI actually drive decisions. So every improvement needs a workaround. Every workaround adds friction. And the product hits a ceiling no amount of prompt-tuning can fix.

Bolt-on AI and AI-native architecture get treated as interchangeable. They produce fundamentally different products.

When AI is the core of how a product works, data flows inform the model continuously. The UX is shaped around what the AI can actually do well. The architecture supports iteration instead of resisting it. When AI is just a feature layer, it's decorating a house built for someone else.

If you're making a build-vs-integrate decision right now, understand the architectural cost before you commit. The gap between "has AI" and "is AI-native" only widens with time 🧠

25/05/2026

Most AI products sit in the same quadrant of this matrix. And it's not the one their teams think.

We built a simple 2x2 model that maps AI integration depth against product dependency. Four quadrants. Four very different realities.

The real question is straightforward: if you removed the AI, would your product collapse or just lose a feature? That answer tells you everything about whether you're truly AI-native or just AI-bolted-on.

We build AI-native production systems every day here in Horten. That hands-on work is what made this framework feel obvious to us.

Swipe through to find where your product sits 💡

23/05/2026

AI-native (adj.): Built with AI as the load-bearing wall, not the wallpaper.

Most companies slap an AI feature onto an existing product and call it "AI-native." That's like putting a skylight on a shed and calling it a greenhouse 🙄

The term has become a marketing checkbox. But it was supposed to mean something specific, and that meaning matters if you're building or buying software in 2026.

We define AI-native as applications where AI is the core of how the product works. Architecture, development, deployment, all shaped around AI from the ground up. Not added as a feature after the fact.

That distinction is practical. A bolted-on AI feature breaks when the underlying product changes. An AI-native architecture adapts, because the AI isn't decorating the system. It IS the system.

So if you're evaluating vendors, try this filter: "If we removed the AI layer, would your product still function?" If the answer is yes, you're looking at wallpaper.

We build this way every day in Horten. And we think the market deserves sharper language for what "AI-native" actually means 💡

21/05/2026

Most teams say they're "building with AI." What they actually mean is they added an API call after the architecture was already decided.

That's bolt-on AI. And it's everywhere.

The tell? Look at when the AI decision entered the project timeline. If it showed up after the data model, after the core logic, after the UX was designed, it's decoration. A feature stitched onto a product that was never built to learn or adapt on its own.

We see this constantly in our consulting work. A team will proudly demo their "AI-powered" product. Then you ask one question: "What happens if you remove the AI layer?" If the answer is "the product still works basically the same," the AI is cosmetic. It's not load-bearing.

AI-native means the product doesn't function without the intelligence. Remove it, and there's nothing left to ship. Bolt-on AI looks impressive in a pitch deck and checks the "AI" box for stakeholders, but it never compounds. It never gets smarter from usage. It just sits there, calling an API, returning a result, disconnected from the product's actual decision-making.

If you're funding an AI project, ask the team one thing: was the AI decision made before or after the architecture was set? That single question separates building with AI from decorating with it 🧠

27/04/2026

The AI on your site already answers. It finishes sentences. It gets the job done.

Does it look like it belongs to you?

Most chat widgets look the same. Same bubble, same generic tone, same nothing-face. They get the words right, sometimes. They never get the room right.

ASMI is a layer on top of what you already built — a face, a voice, and a way to behave that's yours. Composed for a clinic. Warm for a craft brand. Sharp for an agency. The avatar reads the conversation as it happens, and shifts visibly when something shifts in the exchange.

It's a humble first step toward digital experience that finally feels emotionally attuned. More coming.

→ broen.tech/apps/asmi

09/11/2020

Våre venner i Duplo Media gjør det bra :) Heia Bengt Berglund & co!

– I oppstarten lurte vi på om vi fikk lønn neste måned. Det er rart å tenke tilbake på nå.

09/11/2020

https://finansavisen.no/nyheter/teknologi/2020/11/08/7585328/captiva-kjoper-51-prosent-av-broentech-solutions-og-satser-millioner-pa-teknologi-til-vannkraft-og-vindkraft

Captiva satser et tosifret millionbeløp på teknologi til selskaper innen fornybar energi når de investerer i Broentech Solutions. Et utviklerteam på 15 skal jobbe mot vann- og vindkraftbransjen. +

04/11/2020

Gode nyheter, vi har fått en ny majoritets-eier. Vi er klare for å skyte fart og gleder oss over det nye samarbeidet med Captiva.

Som et ledd i sin digitaliseringsstrategi for vannkraftsbransjen kjøper Captiva Digital Solutions AS seg inn i programvareutvikleren Broentech Solutions AS.