National governments are struggling to be strategic in the late 2020s, and many are trapped in short-term reactions rather than shaping events. This briefing note argues that this is not mainly a leadership flaw. It is an institutional design problem in the centre of government. The paper looks at current and past examples of strategy teams in governments and proposes designs for future ones that can combine intelligence, speed and relevance.

The core recommendation is to treat strategy as a function rather than a document. The paper argues that strategy teams must work faster in more uncertain conditions, while keeping direction across multiple timescales. It points to methods that reduce overconfidence and test assumptions, including simulations and games, red teams and pre-mortems. It also highlights a newer method: using multiple artificial intelligence (AI) tools and large language models for scanning, synthesis, modelling and red-teaming, while staying vigilant about error and bias. 

The paper introduces the 'strategy stack'. It frames strategy as layered work that can be distributed, but still needs to be consciously shaped, commissioned and connected.

Strategy stack layers

1. analysis of current data, problems and threats
2. foresight, scenarios and futures
3. experiments and innovation where policy is uncertain
4. policy design that feeds into budgets and laws
5. communications, narratives and frames
6. synthesis that integrates inputs with political priorities for action.