The Faroe Islands’ AI strategy needs a living memory

The question I hear most often about the Faroese AI strategy is a simple one: where are we now?

The short answer is that the pre-study is finished and the real action-plan work is under way. The harder answer is that a national strategy is not one document moving neatly from draft to publication. It is a growing body of evidence, workshop findings, choices, objections, responsibilities and revisions.

If that knowledge cannot be retrieved, the project will repeatedly lose its own context. If it cannot be communicated, the public will only see occasional announcements rather than the work itself.

That is why I increasingly think about the project as two connected systems: an organisational memory in Usable, and a public operating surface at vitliki.talgildu.fo.

Where the work stands

The preparatory work was commissioned in October 2025. Flowcore led the pre-study with Vitlíkisstovan as subcontractor. The assignment was to formulate a vision, gather contributions from institutions and other stakeholders, learn from neighbouring countries and help shape recommendations to the Government.

From November through January, a series of workshops examined six connected areas: skills and understanding, public services, business competitiveness, the Faroese language, technical infrastructure, and ethics. The point was not merely to collect opinions. It was to identify uses, obstacles, priorities and recommendations from people who know the affected parts of society.

The pre-study was completed in March, published online in April and handed to the Minister of Finance in May. Its central direction is now public: the Faroe Islands should show how a small society can use AI responsibly to strengthen its people, language and economy.

On 6 May, around 100 people gathered in Smæran to launch the action-plan phase. The task now is to turn the vision and recommendations into concrete initiatives with ownership and funding, ready for the 2027 budget process. The public site describes the current work as continued stakeholder input, learning from other—especially Nordic—countries, and translating the pre-study into action.

So the project is no longer asking only, “What should the Faroese approach to AI be?” It is asking, “What will we do, who will own it, and how will it be financed?”

What the public site already does well

vitliki.talgildu.fo has become a useful record of the first phase. The pre-study is available as a report and in an English web version. The workshop programme and summaries show which questions were asked and which parts of society were invited into the process. The launch article includes presentations, participation results and the questions raised from the room.

That is more valuable than a polished announcement with no evidence behind it. A reader can follow the movement from workshops to report to action-plan work. The site also communicates the three main pillars clearly—people, language and economy—and keeps international learning visible alongside local participation.

The limitation is freshness. As I write this on 13 August, the latest item in the public news stream is the launch on 6 May. The homepage says that work is happening now, but it does not yet let a reader follow that work in a structured way.

The gap is not necessarily inactivity. It is a lag between the project’s working memory and its published record.

The organisational memory underneath

A project like this produces knowledge faster than it produces reports. There are source documents, workshop notes, research, alternative formulations, meeting outcomes, open questions and reasons why one proposal replaced another. Much of the value lies in those connections.

We now have a dedicated Vitlikisvitan workspace in Usable with curated material from both the pre-study and the main action-plan project. The important word is not storage. It is retrieval.

When a new question appears, a person or an AI agent should be able to find the relevant history: What did the language workshop conclude? Which assumption came from international research? Has an initiative already been proposed? Who challenged it? What evidence changed the direction? Which version is current?

A folder can preserve files. A living organisational memory should help the next person recover the reasoning across them.

That does not make Usable an authority, and it does not mean internal notes should automatically become public. The repository, approved documents, named decision-makers and current project evidence remain the sources of operational truth. Some material is preliminary or sensitive. Some conclusions become stale. Every retrieval still needs provenance, status and human judgement.

The useful loop is therefore:

capture the source → retrieve the relevant context → decide → verify → publish what is approved → write the outcome back

The final step matters. Without it, each workshop, decision or public update becomes another endpoint. With it, the project can accumulate knowledge instead of merely accumulating documents.

From archive to public operating surface

The public site and the internal memory should do different jobs. Usable can hold the detailed evidence and working context. The public site should turn the approved part of that knowledge into something citizens can understand and follow.

In practice, I would like the site to show more than a sequence of news posts. It should say what phase the work is in, when that status was last reviewed and what comes next. Each initiative should eventually have a visible status, owner, next milestone and supporting sources. Important changes should remain traceable instead of silently replacing yesterday’s version. Open questions should be clearly separated from confirmed decisions.

This would make the site less like a campaign page and more like the public home of the strategy: a place to understand the direction, inspect progress and know where participation is still possible.

It would also make communication easier to sustain. A meeting outcome recorded with a clear owner, source and public status can become a short update. A changed initiative can update both its own page and a monthly progress note. A completed pilot can become a case with evidence, safeguards and reusable learning. The publishing process starts from retrievable knowledge rather than from an empty page.

The strategy should practise what it recommends

The Faroese AI strategy will make claims about how institutions should use AI responsibly. The strategy project itself is an opportunity to demonstrate that practice.

AI can help retrieve context across a long-running project, compare new proposals with earlier evidence and prepare clearer public explanations. But it should do so inside a governed process where people can inspect the sources, correct the record and decide what becomes official.

For a small society, continuity matters. We cannot afford to lose important reasoning every time a consultant leaves, a working group changes or a new tool arrives. We also cannot expect public trust if only the final recommendation is visible and the path towards it disappears.

The work now is to produce an ambitious, actionable plan for the 2027 budget. Alongside that, we should build the memory and publishing discipline that allow the plan to keep learning after it is launched.

The public can already see the foundation at vitliki.talgildu.fo. The next step is to make the living work easier to follow—and to make sure the knowledge behind it remains available for the next decision, not trapped in the last meeting or the last AI conversation.