Agentic Planning
Leading-edge empirics and dynamic scenario modeling for industrial strategy and planning.
Good industrial policy requires a detailed picture of technological, productive, and economic systems as an empirical foundation. Until recently, that picture has been too coarse and incomplete to meaningfully inform smart strategic moves. Today, however, artificial intelligence has made it possible to assemble and create highly detailed bespoke datasets for industrial strategy (e.g. the fine-grained value chain decompositions at our first-of-its-kind project valuechains.ai).
With the proliferation of data, however, policymakers can easily get stuck in the analysis phase rather than proceeding swiftly to policy design and implementation.
Agentic Planning solves this challenge.
Our core project is the development of a rigorously structured AI planner that constructs, tests, and explains industrial-policy scenarios, while continuously growing a versioned knowledge base of the productive capabilities and economies of countries and regions across the globe — a systematically connected picture of their technological, productive, and economic systems.
It combines a language model’s generative reach with that ever-growing evidence base, explicit goals and maxims, and independent moral, legal, and institutional checks.
Humans set the orientations and adopt the goals. The agent maps feasible pathways, estimates consequences, and must pass a release gate that blocks unsupported or selectively incomplete proposals.
The whole stack can run on government-controlled infrastructure, so planning does not depend on an external model API. We work with governments and multilaterals to set up the system and run it on their own infrastructure.
In addition to developing the Agentic Planner, we collaborate with governments to help build AI sovereignty.
This includes:
- 01
applying our value chain analysis and agentic planning methodologies to understand, target, and develop countries’ productive capabilities throughout the value chains of AI infrastructure
- 02
working with governments to train local talent to deploy their own open weight AI models
- 03
working with governments to build and deploy their own, independent agentic planning capabilities.