A $2.2 Trillion Fund Showed Exactly How They Do AI. Steal This.

Norway's sovereign wealth fund published their full AI playbook. The lessons start well before the technology.

The Best AI Playbook I’ve Seen Just Got Better. Here’s What You Can Steal From It.

TL;DR: Norway’s $2.2 trillion sovereign wealth fund just showed the world exactly how they do AI. On camera. With real employees. More than half their staff now write code. Their AI team is ten people. They went looking for one big use case and found 171 small ones instead. And they’ve already had to redo their training because the technology moved too fast. This is the best blueprint I’ve seen.

They Showed Their Work. Again.

About a year ago, I wrote about how NBIM saved 213,000 hours annually through AI adoption. I called their approach a model for every professional services firm. I wasn’t wrong.

But last week they did something I almost never see from a major institution. They published a full seminar, on camera, with their employees walking through ten real AI use cases. Not marketing. Not a glossy case study with the rough edges sanded off. Actual people showing actual tools, being honest about what worked and what didn’t.

Watch the whole thing if you can. But I also know you’re running a firm, so here’s what I pulled out of it and what it means for you.

The Foundation Lesson Most People Skip

Their Chief Technology and Operating Officer, Birgitte Bryne, spent her entire segment talking about things that happened before AI. Not the sexy stuff. Insourcing operations from external vendors. Moving to public cloud. And then the part nobody wants to hear about: migrating their databases and cleaning their data.

Here’s what I loved. Tangen himself, on stage, asks her if cleaning data is fun. Her answer: “It’s no fun at all.” He pushes. Does anybody thank you for it? “No.” So how do you get people to do it? She basically said: you set a hard deadline, you tell people the old systems go dark on January 31st, and if you’re sitting there the next day with nothing, that’s on you.

I want you to notice what just happened there. The CEO of a $2.2 trillion fund is on stage, in front of external guests, highlighting the most boring work in the entire organization. Not the AI agents. Not the trading algorithms. The data cleanup. That tells you everything about what actually matters in this stuff.

Because here’s the pattern I see constantly. Firms want to skip straight to the AI part. The chatbots, the agents, the cool demos. But if your data lives in SharePoint folders, email attachments, and individual laptops, the AI doesn’t have anything useful to work with. NBIM got their data house in order first. That’s why everything else worked.

Mandatory Beats Voluntary. Every Time.

CEO Nicolai Tangen was blunt about this. When they rolled out mandatory AI training, people hated it. His words: “It’s like going back to primary school.” But he made it mandatory anyway, because “the people who don’t want to do it are the people who need it the most.”

I’ve been saying a version of this to managing partners for a while now. When AI training is voluntary, you get a predictable result: the 20% who were already curious get better, and the 80% who are skeptical never touch it. The gap widens. That’s the opposite of what you want.

NBIM created seven 30-minute sessions covering prompting, responsible AI, critical thinking with AI outputs. Everyone did them. No exceptions. And here’s the kicker: they’ve already had to run a second round because the first curriculum got stale. AI moved that fast.

That’s not a failure. That’s just what 2026 looks like. Your training program can’t be a one-and-done event. It’s a rolling commitment, and if that sounds exhausting, well, it is. But the alternative is worse.

There Is No Silver Bullet Use Case

This was maybe the most honest thing anyone said in the whole presentation. Stian Kirkeborg, their head of AI, explained that they went hunting for the one big AI use case that would make the fund 20% more efficient. They couldn’t find it. What they found instead was 171 smaller projects.

I actually think that’s good news for most organizations. You don’t need to wait around for some perfect idea. The gains come from a lot of people solving a lot of small problems. A few hours saved here, a manual step eliminated there. It compounds.

Here’s what that looks like in practice. One person automates a financial note that used to take a full week. A two-person comms team builds a media monitoring system that replaces a six-figure vendor contract. A lawyer creates a negotiation simulator that predicts 80% of the other side’s arguments before walking into the room. None of these change the world on their own. But stack 171 of them together and you’ve got a different organization.

Non-Developers Are Building Real Tools

Here’s the stat that stopped me cold: more than half of NBIM’s employees now use Claude Code to build their own solutions. Not software engineers. Portfolio managers. Comms people. Finance controllers. Compliance officers.

Sara Foss from their comms team built an entire AI-powered media monitoring platform called Echo. Sentiment analysis across 50,000 articles a year, data stored in Snowflake, a chatbot that generates reports on demand. Her press team? Two people. She’s not a developer.

Torjus from financial reporting used Claude Code and Cursor to rebuild their quarterly financial statement process from scratch. His team of two automated notes that used to take one person an entire week. Now they take a couple hours.

If you run a law firm, pay attention to this. The future isn’t hiring a bunch of engineers. It’s giving your existing people the tools and permission to build what they need. The person who understands the problem best is usually the one who can solve it fastest, if you get out of their way.

Small Team, Big Impact

NBIM’s AI enablement team started with three people. It’s now ten. For a fund managing $2.2 trillion. And Kirkeborg was clear: his team doesn’t do the AI work. They enable it. Tools, platform, training, support. The building happens everywhere else.

They also created an ambassador network. Twenty volunteers from different departments, each given one job: find the highest-value AI use case in your area and solve it. They got support from the AI team and from Anthropic (who ran twice-weekly training sessions for two months). Then each ambassador showcased results to the rest of the organization. Peer proof, not consultant proof. That’s a different thing entirely.

You probably don’t need a ten-person AI team. Two or three people focused on removing obstacles and keeping momentum going would be a great start. Add a handful of internal champions who can show their colleagues what’s possible, and you’ve got the basics of what NBIM built.

Scrum Is Dead. Move Faster.

This one will ruffle some feathers. NBIM used to run standard Scrum: eight developers, one business person, daily standups, sprint retrospectives, all the ceremonies. They’ve basically thrown that out. Now it’s two developers and one business person, empowered to make decisions and move.

Why? Because AI handles so much of the development work that the old model creates more friction than value. All those rituals designed to keep big teams aligned just slow down small teams that already know what they’re doing.

And this applies way beyond software development. If your firm’s process for trying a new AI tool involves six months and three committees, you’re losing ground to the firm that lets a partner and an associate try something Tuesday and has a working prototype by Friday. Speed matters now in ways it didn’t two years ago.

Governance That Keeps Pace

Here’s where NBIM avoided a trap I see constantly. They built a responsible AI program, but they built it to move with the technology instead of blocking it.

Four layers: a responsible AI guideline aligned with the EU AI Act, an operating model that turns the guideline into actual processes people follow, a cross-functional governance working group that stays current on regulatory changes, and mandatory training for everyone on what AI can and can’t do.

The key phrase from Lydia Gill, their lead privacy and AI governance advisor: a governance structure is only as strong as the people inside it. She’s right. You can write all the policies you want. If your people don’t understand the technology well enough to spot problems, those policies are just paper.

What to Do This Week

1. Audit your data. Not a six-month project. Just answer one question honestly: if you plugged an AI tool into your firm’s data tomorrow, what would it actually have access to, and how clean is it?

2. Make your next AI training session mandatory. Not optional. Not “strongly encouraged.” Mandatory. Keep it short, 30 minutes, and make it practical.

3. Find three to five internal champions who are already curious about AI. Give them a real problem, a tool, and two weeks. Let them show results to the rest of the firm.

4. Stop looking for the one perfect use case. Pick five small ones and start. The compound effect does the heavy lifting.

5. Build governance that says yes with guardrails, not no with exceptions. NBIM proves you can move fast and stay compliant.

The Real Takeaway

What makes NBIM worth studying isn’t any single use case or tool. It’s that they treated AI adoption as a leadership problem, not a technology problem. The CEO pushed and kept pushing. Training was mandatory. The boring data work came first. And then they gave their people room to figure out the rest.

A lot of organizations say they want to be AI-first. NBIM actually did it, on camera, and showed you how. You don’t get that kind of transparency very often. Use it.

---

If you read this far, you’re not wondering whether AI matters. You’re trying to figure out how to get 200 people moving in the same direction without a Norwegian work ethic. That’s the real problem, and it’s a leadership problem more than a technology one.

That’s the conversation I have every day with managing partners, COOs, and practice leaders who are past the hype and into the hard part. If you’re working through what your version of NBIM’s playbook looks like, send me a note at steve@intelligencebyintent.com. Tell me where you’re stuck. I’ll tell you what I’m seeing work and what I’d skip, with no pitch attached.