Back to projects
Economics

Automation Frontier Dashboard

Interactive labor-automation dashboard that explains where AI can automate work, where human bottlenecks remain, and how different roles are exposed.

Next.jsTypeScriptLabor EconomicsAutomationVercelData Visualization

Overview

Automation Frontier is an interactive dashboard for thinking about AI and labor-market change. It ranks industries by automation risk, explains bottlenecks like trust and regulation, and lets users test specific job titles in a sandbox.

The project is built to be understandable without needing an API key or paid model provider. Curated examples use hand-written analysis, and custom jobs fall back to a local rules-based estimator.

Problem

Most automation discourse is either too abstract or too alarmist. A useful tool needs to separate several questions:

  • Can AI perform the knowledge-work layer?
  • Does the job require physical execution?
  • Is there regulation, licensing, or legal liability?
  • Do customers or institutions still require a human?
  • Is the realistic outcome replacement, augmentation, or workflow compression?

The product goal was to make those tradeoffs visible in a dashboard that a non-technical user could scan quickly.

Approach

Bottleneck Index

The main dashboard ranks industries across five dimensions:

DimensionMeaning
AI capabilityHow much of the work current AI can assist or perform
Physical frictionHow much real-world execution slows automation
RegulationHow much law, licensing, audit, or safety review matters
TrustHow much users expect human accountability
Automation riskComposite exposure after capability and bottlenecks

Each row expands into occupation-level examples with plain-language explanations, visible exposure scores, and bottleneck notes.

Automation Sandbox

The sandbox lets users enter a job title and get a structured automation profile across seven dimensions:

  • knowledge work
  • judgment
  • trust
  • physical execution
  • legal liability
  • human preference
  • cost to deploy

The first version used an Anthropic API route for unknown roles. I replaced that hard dependency with a free local analyzer so the feature remains usable in production without paying for inference.

Local Analysis Fallback

The local fallback maps a job title to the closest industry, pulls related occupation signals, scores the role, and returns a readable summary with a confidence label.

This makes the product more robust:

  • no API key required for users
  • no runtime failure when model calls fail
  • deterministic outputs for common searches
  • optional AI enrichment can be added later without blocking the core feature

Tech Stack

  • Next.js App Router
  • TypeScript for scoring types and API contracts
  • Tailwind CSS for the dashboard UI
  • Vercel for deployment
  • Local JSON datasets for industry and occupation signals
  • Optional Vercel AI SDK path for future model-backed analysis

Results

The deployed dashboard supports:

  • industry-level automation ranking
  • hover explanations for metrics
  • expanded occupation examples
  • local job-title analysis without paid model APIs
  • confidence/source labels for sandbox outputs
  • a case-study ready interface that explains what users are seeing

What This Proves

  • I can turn an economics concept into an interactive product.
  • I can design around user comprehension, not just data display.
  • I can remove paid API dependencies while preserving core product behavior.
  • I can build practical AI-adjacent tools that degrade gracefully.

Next Steps

  • Add more occupation data from BLS/O*NET.
  • Add confidence bands and source citations per score.
  • Let users compare two roles side by side.
  • Add downloadable summaries for career planning or workforce analysis.