Innsikt om AI governance, implementering og digitalisering for norske virksomheter.
We write night watch plans: instructions that let Claude work for long stretches without supervision. Here are the rules we have learned, what the logs show, and what research says about agents that grade their own work.
Jev is a new kind of model that answers with a decision and a number for how sure it is, not with text. It may be the missing piece in the AI loop. But who decides what is sure enough?
September produced three major AI results. None of them came from training the model on the problem. They came from the loop around the model, and that difference matters for ordinary businesses.
There's a divide running through Norwegian working life right now. Some use Claude and ChatGPT as a better search engine. Others have moved the work itself into them. The two groups see completely different futures.
A quality management engine that checks an accounting firm's documentation against the Norwegian control form. The control points live as text files, not as code, which is why the same machinery can point at other reporting-heavy industries.
Anthropic now watermarks all text Claude generates - even when it just shortens something you wrote yourself. The law actually has an exemption for exactly this. Claude doesn't use it.
The AI literacy requirement in the EU AI Act becomes enforceable on 3 August 2026. Here's the model we recommend to actually meet it: a cross-functional AI committee with one owner per department.
IT Buddy has been named an OpenAI Select Partner within the OpenAI Partner Network. Here's what that actually involves – and why it doesn't change which tool we recommend for you.
Claude Cowork has a new feature: screen-record yourself doing a task, and it automatically becomes a reusable Skill. Here's how it actually works, with examples across industries – and why the governance side of it still hasn't caught up.
AI agents are often treated as extra users with far too much access. Microsoft Entra Agent ID and Access Packages fix that with time-bound, approved access – and that's exactly why agents can actually be safe to run in production.
Kimi K3 is the largest open-weight AI model ever released, and the price war it's triggering is reshaping what AI costs. But it also raises a harder question: is American AI actually safer than Chinese AI – or do we trust it out of habit?
A guide to using Claude professionally: pick the right entry point, set up Projects, Skills and Memory correctly, and choose the right model tier for the task – not out of habit.
Anthropic guarantees at least 60 days' notice before retiring a model. OpenAI can give as little as two weeks on preview models. The answer isn't to leave the major providers. It's to build redundancy.
Anthropic has launched Claude Fable 5, its most capable, widely released model yet. It costs twice as much as Opus 4.8. Here's when that's actually worth it.
Claude can handle IT support in Microsoft 365. The question is: as whom, with what access? Here's how we build it safely, with a bot that has its own identity.
Most MCP servers only implement tools. The spec defines five primitives. Here's what prompts, resources, sampling, and roots actually do, and why they matter.
You've heard Claude can now connect to your business systems. MCP is the mechanism that makes it possible, and the thing that determines whether it's safe. Explained without the jargon.
Your AI assistant is probably already saving time. But productivity is step one of five on the maturity ladder, here are the next four, and how to climb them.
Most businesses already own the tools to detect and govern shadow AI inside Microsoft 365, they've just never switched them on. Here's how, in four weeks.
Most businesses have tried Claude or ChatGPT in the browser. It sits next to the work, not inside it. Here's what it takes to close that gap, and why it starts with structure.
Anthropic released Claude Opus 4.8 just 41 days after the previous version. It's more capable and more honest, at the same price. But for Norwegian businesses, the model choice isn't what decides whether AI delivers value. Structure is.
Anthropic has made Claude Cowork generally available, an agent that works directly with your files, Google Drive, Gmail, and DocuSign. Here's what Norwegian businesses need in place before turning it loose.
Since March 2026, Claude remembers what users tell it, by default, including on the free plan. For Norwegian businesses, that means employees are building a private memory bank of work information, outside the company's control.
Anthropic launched its most powerful model – Claude Mythos – but is holding it back because it's too good at finding security vulnerabilities. Here's what it means for Norwegian businesses.
Anthropic launched Claude Managed Agents in public beta on April 8, 2026. Here's what it is, what it means for Norwegian businesses – and what you need in place before you can use it.
Claude or ChatGPT – which should Norwegian businesses choose? We compare data handling, GDPR compliance and admin controls – not just AI performance.
Microsoft Copilot is disappointing many Norwegian businesses. Here are the three most common reasons – and what you actually need in place before paying for the licence.
The EU AI Act classifies AI in recruitment as high-risk from August 2026. Here is what Norwegian businesses need to have in place, and what you can safely automate.
Want to implement AI in your business but don't know where to start? Here is the practical step-by-step guide for Norwegian businesses.
Your company's best methodology lives inside the heads of key people. That's a risk. Here's how to document it, and make it available to AI agents.
The EU AI Act is coming to Norway through the EEA. Here is what it actually means for you as a Norwegian business owner – without the legal jargon.
An AI policy is the single most important step a business can take for responsible AI use. Here is what it should include – and a concrete template to get you started.
Many employees use ChatGPT at work without management knowing. Here are the 5 biggest risks – and what you can do to manage them.
The EU AI Act is Europe's landmark law on artificial intelligence. We explain what it actually requires, who it applies to, and what Norwegian businesses should do now.
A walkthrough of Claude Enterprise – Anthropic's business version of Claude AI. What sets it apart from the free version, and is it the right choice for Norwegian businesses?
A practical introduction to AI governance for Norwegian businesses – what it is, why it matters, and what you should do now.