See how a policy plays out before you commit to it.
When automation displaces workers, PolicySimAI projects how they respond — who retrains, who searches for similar work, who exits, who takes a job below their skill level — and shows how retraining subsidies and Universal Basic Income shift the outcome. Every projection is calibrated to UK official statistics and validated against real displacement episodes.
Calibrated to UK official statistics · Validated on Redcar, Port Talbot & the Hartz reforms
From Scenario to Decision
Simulate, compare, and commit — all in one workflow.
1
Describe the scenario
Tell PolicySimAI the shock you're facing — a wave of automation displacing a group of workers — and the levers you're weighing, like a retraining subsidy or UBI.
2
Run the simulation
A calibrated choice model projects how each type of worker responds — retrain, seek similar work, exit, or accept underemployment — across a synthetic population.
3
Compare and decide
Read the outcome breakdown and stakeholder trade-offs, test alternative levers side by side, and take a defensible brief into the room.
Simulations You Can Trust
Every run ends in a decision-ready brief — grounded in a statistical core, not a black box.
Policy Briefs
Three or four sentences a minister can act on — synthesised from the outcome breakdown, per-archetype behaviour, and a structured dialogue between stakeholder perspectives.
AI layer — explains
Narrates, flags anomalies, weighs stakeholder trade-offs.
Statistical core — decides
A calibrated choice model runs every per-tick projection.
The AI explains. The economics decides.
The language model never touches the outcome math. Every number traces back to a validated, versioned model — reproducible and defensible to any economist on your team.
Built on Peer-Reviewed Research
The engine is a calibrated discrete-choice model — the standard framework in labour economics — fit to published data and checked against displacement events that actually happened.
3 episodes
Validated against real displacement
Redcar, Port Talbot, and the German Hartz reforms — cross-validated leave-one-episode-out.
RMSE ≤ 0.10
On held-out historical outcomes
The out-of-sample error is printed on the cover of every model — no accuracy claim without a number.
Every model versioned
Auditable, not verbal
Coefficients, the episodes that fit them, and the validation score ship as versioned files in source control.
Methodology reviewed with researchers at
Policy Domains
PolicySimAI covers a wide range of policy areas, from labour markets to fiscal policy.
Labour Markets
Automation displacement, retraining, and workforce participation — calibrated and validated today.
Housing · Healthcare · Climate
The engine is scenario-agnostic. New domains plug in as a new calibration, not a rebuild.
Mission & Vision
Transforming policy analysis with AI — so decisions can be tested before they’re made.
Our Mission
We help institutions test the likely effects of policy decisions before real-world rollout — combining AI, structured scenario analysis, and digital twin thinking so governments, researchers, and organisations can explore economic, fiscal, and social outcomes with greater speed, transparency, and accountability.
Our aim is to make policy testing more accessible, evidence-led, and auditable, so decision-makers can assess trade-offs and act with greater confidence.
Policy, business, academic, and AI specialists.
A team built for high-stakes decisions.
PolicySimAI is made up of a team of policy, business, academic, and AI specialists who have worked with public institutions on complex, high-stakes decisions.