The Man Who Wrote Iceland's first AI Playbook — and Why Data Governance Is the Real Battle

When spreadsheets arrived in 1996, every office in the world had to buy a computer. It was not a gentle transition. It was a binary shift: either you adapted, or you fell behind. There was no middle ground, no wait-and-see strategy that ended well. I have been thinking about that analogy a lot since sitting down with Gisli Ragnar Gudmundsson for this episode of Taming Technology, because he believes artificial intelligence represents that same kind of inflection point — except this time, the stakes are higher and the clock is moving faster.

Gisli is one of those rare people whose career trajectory reads like a novel. An industrial engineer who spent years designing hearing devices in Germany, the UK, and China, he returned to Iceland and ended up at Rikiskaup — the State Procurement agency — where he was supposed to help public institutions adopt technology. Then, almost by accident, he was pulled into the Ministry of Higher Education and Innovation, where he spent nine months helping craft Iceland's AI Action Plan to 2026. He has since moved to KPMG, but the insights he carries from that government work are, frankly, sobering.

The Data Governance Crisis Nobody Wants to Talk About

Here is a number that should keep public administrators awake at night: Iceland has over 160 government institutions, each managing its own data with no unified standard. Some are still running on-premise servers — literally machines humming in basement rooms. When Gisli arrived at Rikiskaup, he discovered that even requesting the government's own procurement data took four to five months, not because of bureaucratic obstruction, but because no one had ever anticipated that someone would actually ask for it.

This is not a technology problem. It is an infrastructure problem. Gisli, with his engineering mind, put it in terms I found immediately clarifying: data should be treated like any other national infrastructure — like roads, electricity, or plumbing. You need blueprints. You need standards. You need someone who knows where the pipes run. Right now, Iceland does not have that. Multiple institutions sit on the same data — ID numbers, addresses, contact information — duplicated across systems, with no central oversight and no data architect to draw the map.

"If the data is not ready, and you pour AI on top of it, you are not solving a problem — you are creating a much more sophisticated mess."

He told me about a bank in Iceland that hired twenty to twenty-five data scientists and spent three to four years simply cleaning and organizing their data before they could do anything meaningful with AI. That timeline is instructive. The hype cycle suggests you can deploy a large language model and transform your organization overnight. The reality, as Gisli puts it, is far more mundane and far more essential: you have to do the plumbing first.

What Singapore Understood in 2017 That Iceland Is Still Figuring Out

When Gisli mentioned Singapore, I sat up a little straighter. Since 2017, Singapore has given every citizen roughly 500 dollars in credit to access digital training platforms — structured courses, certifications, practical skills. It is not an experiment; it is national infrastructure for human capability. The government decided that digital literacy was not optional and that leaving citizens to figure it out on their own was not a strategy but an abdication of responsibility.

Iceland, with its 380,000 people and strong welfare traditions, could do something similar — arguably more easily. We have powerful trade unions with education and retraining funds. We have a small enough population that systemic change does not require moving mountains. But as Gisli pointed out, the conversation between government and labor unions had only just begun when political disruptions — dissolved governments, new ministers, reorganized ministries — put everything on pause. The AI Action Plan was published, and then the momentum stalled.

The Generalist's Revenge

One of the most resonant moments in our conversation came when Gisli referenced the book "Range" by David Epstein, which argues that generalists — people with broad, cross-disciplinary knowledge — are better equipped to solve the complex, novel problems of the future. Gisli calls himself a jack of all trades, master of none, or perhaps better, a master of one. I know that feeling intimately. My own path — from paramedic and firefighter to AI project manager — hardly follows a straight line. But it turns out that the diagnostic thinking I learned in emergency services, the ability to triage, to stay calm under uncertainty, to synthesize information from multiple sources — these are precisely the skills that matter when you are trying to help an organization navigate AI adoption.

Gisli and I agreed: the ideal team for this era combines generalists who can communicate, empathize, and see connections across domains with hyper-specialists who go deep in their field. For a nation as small as Iceland, this is not abstract organizational theory. It is survival strategy. We do not have the population to produce enough specialists in every emerging field. What we can produce is adaptable, curious people who know how to learn — and who have access to AI tools that amplify their range.

Twenty Percent Education, Eighty Percent Training

Gisli said something that has stayed with me: AI adoption is twenty percent technology and eighty percent training and cultural change. You cannot hand someone a tool and a manual and expect transformation. People need to become their own craftsmen with the technology — to discover how it extends their existing skills, interests, and workflows. Otherwise, as he put it, you are just shoveling coal into someone else's furnace.

He shared a story about his wife, a primary school teacher, that perfectly illustrated this. Rather than sitting her down for a formal course, he set her up with a ChatGPT subscription and built her a custom GPT called "Novo" — named after their old dog. He gave her a basic framework: no sensitive information, just start prompting and ask me when you get stuck. Within weeks, she was not only proficient but genuinely excited, recognizing that the tool helped her in ways she had not expected. That is the model: not top-down instruction, but scaffolded exploration. Democratizing the technology by meeting people where they are.

We also discussed the troubling findings from a Swiss study suggesting that the more people use generative AI, the more their critical thinking skills decline. Gisli mentioned the Dunning-Kruger effect — the famous case of the bank robber who smeared lemon juice on his face, convinced it would make him invisible to security cameras. He was not unintelligent. He was simply so thoroughly convinced of his own reasoning that he never questioned it. Most people, Gisli observed, dramatically overestimate their own critical thinking abilities. When he surveys audiences before his talks, self-assessments of critical thinking are always through the roof. That gap between confidence and competence is exactly where AI can do its most subtle damage — or, if we are intentional about it, its most profound good.

The Safety Net That Already Exists — and the Silence That Surrounds It

Perhaps the most provocative part of our conversation concerned something Gisli explored by feeding Icelandic legislation into ChatGPT. A former minister had proposed changes to disability benefit laws, allowing people to work more without losing their benefits. Gisli asked GPT a simple question: could this be the beginning of a universal basic income? GPT initially said no — these are disability benefits, not UBI. But when Gisli pushed back, pointing out that Nordic welfare states already function as a kind of proto-safety-net, that disability rolls have swelled partly because the labor market is burning people out, that the proposed reforms would let people earn while maintaining a baseline — GPT reconsidered. Yes, it said. In welfare states like Iceland, UBI might not arrive as a revolutionary policy. It might emerge gradually, evolving from existing systems, almost without anyone noticing.

What troubled us both was the political silence. During the most recent Icelandic elections, AI's impact on the labor market was essentially absent from the debate. Candidates discussed the welfare system, healthcare, housing — but the elephant behind the curtain, the technological shift that could fundamentally reshape which jobs exist and who can do them, went unmentioned. As someone who spent sixteen years in emergency services, I know what happens when you ignore warning signs. You do not get to choose when the crisis arrives. You only get to choose how prepared you are.

From Hourly Billing to Human Value

Gisli raised a point about consulting and legal work that I think extends far beyond those professions. When a task that used to take four hours can be done in twenty minutes with AI, the old model of billing by the hour collapses. The shift is toward value-based pricing — paying for outcomes, not time. Solo practitioners who master AI tools will be able to compete with large firms. The question is whether organizations will recognize this shift proactively or be forced into it by competitors who already have.

But there is a human dimension here too. If productivity doubles, do we simply demand twice the output from people in the same eight-hour day? Or do we let people work five focused hours and reclaim the rest for the things that actually make life worth living — conversation, family, rest, thought? Gisli described his own practice of working in two forty-five-minute hyper-focus blocks each morning, then deliberately stepping away for meetings, conversations, and unstructured thinking. It is a discipline, not a luxury. And it points toward what I believe will be one of the defining questions of the AI era: will we use this technology to become more productive, or more human?

Building the Future With Intention

Walking away from this conversation, I felt something I have come to recognize as a recurring theme on this podcast: a mixture of urgency and hope. Urgency because the gaps — in data governance, in political discourse, in workforce preparation — are real and widening. Hope because people like Gisli exist: people who combine international experience with local commitment, engineering precision with policy vision, and a willingness to say uncomfortable truths in rooms that prefer comfortable silences.

Iceland does not need to become Singapore. But it needs to become intentional. It needs data architects and digital training credits and politicians who are willing to say the word "AI" during election season. It needs generalists who can bridge disciplines and specialists who can go deep. It needs teachers like Gisli's wife, discovering the technology on their own terms, and policymakers like Gisli himself, who understand that the action plan is only as good as the follow-through.

The spreadsheet moment is here. The question is not whether Iceland will adopt AI. The question is whether we will do it with the same thoughtfulness and care that built the welfare state in the first place — or whether we will let it happen to us while we argue about something else.

Listen to the full conversation with Gisli Ragnar Gudmundsson on Spotify.