Research. Convene. Build.

There is no shortage of AI principles. There is a severe shortage of the operational artefacts that would make those principles binding: the contract clause, the curriculum module, the disclosure schedule, the review process. Principles are written by one set of institutions and implemented by another, and almost nothing carries the intent across the gap.

ImpactLab exists to produce the missing artefacts, in public, with named contributors, for institutions that can use them without asking anyone's permission. We are a think-do-tank in the literal sense: the analysis is only finished when something usable comes out of it.

Our research is applied, footnoted, and free to read. It starts from a decision somebody has to make — a tender to write, a grid allocation to justify, a screening methodology to defend — and works backwards to the evidence that decision requires.

Three lines are active. Procurement and the clause gap, examining what public AI contracts actually contain against what policy requires of them. Literacy as a policy instrument, tracking the widening adoption gap between the Global North and South and what binding literacy obligations mean in practice. And compute, energy, and social licence, on how clean-grid jurisdictions should evaluate the allocation of power to data centres.

The Dispatch is where this work is published as it develops. Every issue states a bear case: the condition under which we would consider the argument wrong. Longer working papers are published with named civil-society, municipal, or academic co-authors.

We bring named voices to the same table: ministers and machine-learning engineers, regulators and founders, civil-society leaders and capital allocators. Rooms are small, between ten and thirty people, off the record by default, and organised around a single concrete question with a defensible artefact at the end of it.

A convening that produces only goodwill has failed. Each Lab is scoped so that the participants leave holding a draft — clause language, a curriculum outline, a disclosure template — and a named owner for the next version of it.

Recent convenings include Co-Chair of the 26th Harvard Social Enterprise Conference, a keynote at TOMRA's annual gathering on AI and industrial transformation, and a NYC AI Nexus preview at Civic Hall in front of more than 300 founders, investors, researchers, and partners.

Every ImpactLab engagement ends with something a participant can implement the week they return — a vendor due-diligence checklist, a model-risk taxonomy, a regional rollout plan, a curriculum module — not a sixty-page PDF that sits on a shelf.

Active build tracks are the Nordic AI Blueprint, a versioned open clause library for ministries and agencies operating under the EU AI Act and EEA-equivalent frameworks; the AI for Norway curriculum framework, Norwegian-language AI literacy built around the decisions public servants actually make; and sector governance templates for public administration, healthcare, and SME deployment.

Everything we build is published free, versioned, and open to contribution. Contributing institutions are named.

Labs. Closed-door working rooms around one question, with a draft artefact as the exit condition.

Publications. The Dispatch and working papers, footnoted to primary sources, with a stated bear case and a right of reply for any institution named.

Programmes. Nordic AI Summit, NYC AI Nexus, AI for Norway, and the Middle East ATLAS — recurring platforms that keep the same people in durable contact between Labs.

Blueprints. The open, versioned artefacts themselves: clause libraries, curricula, disclosure methodologies. This is the mode the other three exist to feed.

We work across the Nordics, the United States, and the Middle East, with a standing analytical track on Norway and Pakistan as a small-state and large-state analogue in AI literacy.

We do not build models, sell software, or certify vendors. We do not accept funding conditioned on a conclusion or on review rights before publication. We do not name a partner where nothing has been agreed in writing. How that is enforced is set out on the governance page.