FutureGrid

An interactive data platform visualising how AI is reshaping the global workforce — built on current, authoritative data sources, not Frey & Osborne (2013).

Live: futuregrid.genisisiq.com

What It Is

FutureGrid maps AI's real impact on careers, sectors, and countries using observed Claude usage patterns (Anthropic Economic Index), live BLS employment and wage data, O*NET skills, and global AI diffusion and preparedness indices. Every metric is sourced from post-2023 data and is fully attributed — see the in-app Sources page for details.

Features

Page	What you get
Dashboard	Workforce-weighted AI-exposure score, headline stats, and top-impacted occupation highlights
Careers	Search, filter, sort, and compare 750+ SOC occupations by AI exposure, employment, wages, and growth outlook; click any occupation for a full detail page
Sectors	Sector-level breakdowns of AI exposure and employment concentration; drill into individual sectors
Skills	Top skills by occupation and AI-exposure band; reskilling pathway suggestions
Heatmap	Cross-tab heat map of exposure × sector × employment
Global Map	World choropleth with three switchable metric layers — Claude usage index (Anthropic AEI), GenAI diffusion % (Microsoft AIEI, 147 economies including China), and IMF AI Preparedness composite score; click any country to drill down into the four AIPI sub-indices (Digital Infrastructure, Human Capital, Innovation & Integration, Regulation & Ethics)
Sources	Full data provenance, licenses, and caveats for every dataset
⌘K palette	Keyboard-driven command palette for fast navigation

Data Sources

All snapshots are committed to the repository — the app build is fully offline (no API calls at build time unless you explicitly refresh the snapshots).

License & redistribution compliance: See data/COMPLIANCE.md for the full per-dataset audit matrix, /sources attribution gaps fixed, and the remediation plan for flagged non-redistributable items. Issue #57 (bulk-download) is gated on that document.

Dataset	Publisher	License	Used for
Anthropic Economic Index	Anthropic (2025)	CC-BY 4.0	Primary AI-exposure metric (observed_exposure) per SOC occupation; replaces Frey-Osborne 2013. Also provides salary fallback, job forecast, job zone, Bright Outlook, and sector.
BLS OEWS via BLS Public Data API	U.S. Bureau of Labor Statistics (2025)	Public Domain	Per-occupation employment totals and median wages (authoritative; overrides AEI bundle when BLS_API_KEY is set)
O*NET 28.3 Skills Database	USDOL/ETA (2024)	CC BY 4.0	Top skills by importance per SOC occupation
Microsoft AI Diffusion Report — AIEI Q1 2026	Microsoft (2026)	MIT	GenAI diffusion % of working-age population, three-period trend (H1 2025 / H2 2025 / Q1 2026), 147 economies including China
IMF AI Preparedness Index (AIPI)	IMF (2023)	IMF terms	Overall AI readiness composite (0–1, 178 countries) and four sub-indices: Digital Infrastructure (DI), Human Capital & Labour Market Policies (HCLMP), Innovation & Economic Integration (IEI), Regulation & Ethics (RE)
Natural Earth / world-atlas@2	Natural Earth / Mike Bostock	Public Domain / ISC	110 m country polygons (TopoJSON) for the world map choropleth

Comparability note: usageIndex (Anthropic Claude API sessions), diffusionPct (Microsoft behavior-based survey), and aiReadiness (IMF infrastructure score) measure different things and must not be merged or averaged. The Global Map keeps them on separate layers with clear labelling.

Getting Started

Prerequisites

• Node.js ≥ 18
• npm ≥ 9

Install & Run

npm install
npm run dev          # http://localhost:3000

Production Build

npm run build        # static export → out/

The output in out/ is a fully static site ready for any CDN or static host.

Tests

npm run test:run     # vitest — 73+ tests

GitHub Pages

Pushes to main trigger the Deploy GitHub Pages workflow. Set the Pages source to GitHub Actions in repository settings. The custom domain is futuregrid.genisisiq.com.

Set the NEXT_PUBLIC_GA_ID repository secret to enable Google Analytics.

Refreshing Data Snapshots

Snapshots are committed; you only need to re-run these if you want fresher data.

Full occupation + career data

npm run build:data

Runs build:proxies (supplemental usage proxies) then rebuilds the main occupation snapshot. Reads .env / .env.local via Next's env loader.

Required key:

Variable	Where to get it	Used for
BLS_API_KEY	Free — bls.gov/developers	BLS OEWS employment and median wage data (authoritative figures)

Optional key:

Variable	Used for
CENSUS_API_KEY	Supplemental U.S. employer-firm AI adoption data from Census ABS (silently skipped if not set)

Global AI metrics (country-level)

npm run build:global-metrics

Refreshes the country-level data used by the Global Map (Anthropic usage index, Microsoft diffusion, IMF AIPI).

O*NET occupation enrichment

To refresh O*NET occupation enrichment for career detail pages:

npm run build:onet

Set ONET_API_KEY and optionally ONET_ENRICH_LIMIT in .env first. ONET_ENRICH_LIMIT=0 enriches all occupations; the default enriches a priority subset.

World map geometry

npm run build:geo

Regenerates the TopoJSON country polygon file from world-atlas.

Social share image

npm run build:og

Regenerates the branded OpenGraph image (public/og.png).

WARN notices (Wisconsin via Google Sheets)

npm run build:warn

Fetches Wisconsin WARN-act notices from a Google Sheets spreadsheet.

Required environment variable:

Variable	Where to get it	Used for
GOOGLE_SHEETS_API_KEY	Google Cloud Console — create a restricted Sheets API key	Authenticates the Google Sheets API call to pull WI WARN data

Set this as a CI/build secret (e.g., a GitHub Actions secret). Never commit it to the repository. The script will fail with a clear error message if the variable is absent.

Data Provenance

Every data/*.json snapshot carries a standardized meta block so consumers can show when data was refreshed and where it came from:

"meta": {
  "generatedAt": "2026-07-02T08:41:00.816Z", // ISO timestamp the file was produced
  "asOf": "2025",                             // the period the data describes
  "source": { "name": "…", "publisher": "…", "url": "…" }, // or a string
  "version": "1.0.0"                          // meta-contract version
}

The bare-array datasets (occupation-snapshot, occupation-snapshot-slim, country-exposure) are wrapped as { meta, data: [...] }; all other object datasets are annotated additively (existing top-level fields are preserved).

npm run build:provenance scans every snapshot and writes data/provenance.json — a central registry { generatedAt, datasets: [...] } listing each dataset's generatedAt, asOf, source, version, and row count. It is validated (validateProvenance) before writing and runs at the end of the build:data chain. The typed loader lib/provenance.ts exposes the dataset list plus getDataAsOf(id), getDatasetProvenance(id), and getLatestAsOf() helpers for the UI provenance badge and methodology changelog.

CI guardrails

Two extra guardrails run after every build to catch performance and accessibility regressions.

Bundle-size budget (npm run check:bundle)

Scans out/_next/static/chunks/*.js and fails if any single client JS chunk exceeds 700 KB (today's largest is ≈ 503 KB — the budget gives ~40 % headroom while catching the 3 MB-class regressions that were fixed in #47).

Run locally:

npm run build
npm run check:bundle

Accessibility audit (npm run check:a11y)

Serves the static export on a local port, loads each representative page in headless Chrome via Chrome DevTools Protocol, injects axe-core, and asserts zero critical and zero serious violations (the color-contrast rule is temporarily excluded pending a design-system palette update — tracked as a follow-up) across:

Page	Path
Home	/
Careers	/careers
Global	/global
Labor	/labor
Frontier	/frontier
Analysis	/analysis

Run locally:

npm run build
npm run check:a11y

No Playwright/Puppeteer dependency — uses system Chrome (/usr/bin/google-chrome) via CDP; override with CHROME_BIN. Both scripts are Node 20 compatible.

Scheduled data refresh

.github/workflows/refresh-data.yml refreshes the committed snapshots on a weekly schedule (0 6 * * 1, i.e. Mondays 06:00 UTC) and can also be triggered manually via workflow_dispatch. It runs npm ci then the data builds (build:warn, build:state-labor, build:state-qcew, build:jolts, build:data, build:snapshot-slim, build:warn-public, build:provenance) and opens a pull request (peter-evans/create-pull-request@v6, branch data/scheduled-refresh) titled chore(data): scheduled snapshot refresh with any changed files.

The build steps are guarded by the issue #49 sanity gates, so a bad upstream fetch throws and fails the job — that failure is the alert.

Required secret:

Secret	Where to get it	Used for
GOOGLE_SHEETS_API_KEY	Google Cloud Console — restricted Sheets API key	Passed to build:warn to pull WI WARN data from Google Sheets

Tech Stack

• Framework: Next.js 16 / React 19 (static export)
• Visualisation: D3.js 7, Chart.js 4
• Styling: Tailwind CSS 4
• Testing: Vitest 4 + Testing Library

License

MIT