PolicyGrid generates auditable macroeconomic country briefs from live IMF, World Bank, UN and OECD data. Source, timestamp, vintage, and status are attached to every observation, and code — not the AI — does every calculation.
Development economists and donor analysts spend days manually pulling and reconciling data from IMF, World Bank, UN and OECD sources — resolving disagreements between datasets, building charts, documenting every figure for a credible report.
Generic AI tools can now produce this analysis in seconds. But they cannot prove where the numbers came from, which data release they belong to, or whether a figure is an actual observation or a model projection. For policy decisions, that gap is not tolerable.
"The biggest surprise was that the bottleneck isn't access to data — it's trust. Users are only comfortable with AI-generated analysis if they can trace every number back to a source and know whether it's observed or projected."
Founder, PolicyGrid
Every number PolicyGrid produces carries four pieces of provenance, encoded at the observation level — not claimed in a footnote, but attached to each row of data.
The API endpoint, not just the institution name. Every observation records where it came from precisely enough to be verified independently.
Data refreshes continuously. Every observation records the exact UTC pull time and the specific data release it came from — so you know if you're looking at the April or October revision.
Actuals and projections are never silently blended. The status is encoded in the data at ingestion, so a report cannot accidentally present a projection as an observed value.
When IMF and World Bank report different figures for the same indicator and period, both are shown alongside the gap — not silently resolved to whichever number was retrieved last.
These are actual data points stored in PolicyGrid's database — each with the source, vintage, and status that accompany it everywhere the platform uses it.
| Country | Indicator | Value | Vintage | Source | Status |
|---|---|---|---|---|---|
| Kenya | GDP Growth · 2025 | 4.6 % | WEO Oct 2026 | IMF | Actual |
| Zambia | External Debt (% GNI) · 2024 | 114.2 % | WDI Oct 2026 | World Bank | Reported |
| Nigeria | Govt Debt (% GDP) · 2025 | 36.1 % | WEO Oct 2026 | IMF | Actual |
| Pakistan | Tax revenue (% GDP) · 2024 | 12.3 % | Revenue Statistics | OECD | Reported |
| Ghana | CPI Inflation · 2023 ⚠ | 39.2% / 38.1% | Oct 2026 | IMF ↕ World Bank | Conflict |
| Tanzania | Fiscal Balance (% GDP) · 2025 | −2.6 % | WEO Oct 2026 | IMF | Estimated |
| Sri Lanka | Current Account (% GDP) · 2025 | +1.6 % | WEO Oct 2026 | IMF | Actual |
Select a country, configure the scope, and PolicyGrid generates a complete macroeconomic assessment — with charts, source-conflict analysis, an executive summary, and a plain-English one-pager. Every figure in the output is linked to its exact source observation.
Delivered as PDF, Word, or CSV. All formats carry the provenance metadata throughout.
Interrogate live country data or a specific generated brief through a conversational interface. Every answer carries inline citations with source, vintage, and status — the same provenance standard as the Field Brief, delivered in dialogue.
Ask follow-ups, drill into specific indicators, compare countries, or request explanations of data conflicts. Export the full conversation as a sourced PDF.
This is an early MVP. The scope is specific because specificity is a strength with this audience — not a liability to hide.
PolicyGrid doesn't have invented testimonials. The two assessments below are real, from named experts who reviewed it directly.
"The empirical rigour here is exactly what the field needs. Development economists working across South Asia and Sub-Saharan Africa spend weeks doing manually what this does in minutes — and with every figure grounded in a verifiable source, not AI conjecture."
"PolicyGrid takes the work a step further — from making development finance data accessible, to producing auditable, policy-oriented analysis on top of it. The approach is very impressive."
"ONE Data helps you follow the money. PolicyGrid helps you understand the country that money is going into."
Positioning — PolicyGridPolicyGrid runs automated ingestion pipelines against IMF, World Bank, UN SDG and OECD endpoints on a continuous basis. Every observation records the pull timestamp alongside the data vintage — so you always know exactly which release a figure came from, and when it was retrieved.
Actual means official data in the source's release. Estimated means a staff estimate for a past year, and projected means a forecast — both flagged in the source data and in PolicyGrid's output. Where a source publishes no flag (World Bank WDI), the value is labelled reported rather than guessed. The status is encoded at the row level, not inferred from context.
Currently 20 countries: 12 in Sub-Saharan Africa (Nigeria, Kenya, Ethiopia, Tanzania, Ghana, Uganda, Rwanda, Senegal, Côte d'Ivoire, Mozambique, Zambia, South Africa) and 8 in South Asia (Pakistan, India, Bangladesh, Sri Lanka, Nepal, Maldives, Bhutan, Afghanistan). Additional countries and sources are planned for subsequent phases.
General-purpose AI assistants and search tools produce fluent text but cannot guarantee that numbers are current, sourced, or independently verifiable. They silently blend actuals with projections, and their outputs cannot be audited or reproduced. PolicyGrid queries live data sources directly, stores every observation with its provenance, and never lets the model type a number: figures and calculations come from code, and an evidence check reads every sentence before release, removing claims the data do not support.
Field Briefs are delivered as PDF (full report and executive one-pager separately), Word (.docx), and CSV. Grounded Chat sessions export as a sourced PDF. All formats preserve source attribution and data vintage in the document metadata.
PolicyGrid is in active early access. Request access to generate a Field Brief or run a Grounded Chat session on any of our 20 covered countries.
BOOTSTRAPPED · EARLY STAGE · 20 COUNTRIES · IMF · WORLD BANK · UN · OECD