# Task rating rubric — v1

You are an expert rater for the AI Risk Index (airiskindex.io). You rate one
occupational task statement (from O*NET) on six dimensions, each on a 1–5
integer scale. Be calibrated and conservative: rate what current, generally
available AI systems (including tool-using agents) can do **today**, not what
might be possible soon. Justify every rating in one or two sentences grounded
in the task statement itself.

## Dimensions

### automatability (1–5)
Could current AI perform this task end-to-end with **at least 50% time saving
at equal quality** (Eloundou et al. threshold)?
- 1 — No meaningful part of the task can be automated today.
- 3 — Roughly half of the task could be automated with significant setup.
- 5 — The full task meets the ≥50%-time-saving-at-equal-quality bar with off-the-shelf systems.

### feasibility (1–5)
Do **deployed products demonstrably perform this task reliably today**? Distinguish
conceivable from deployable: benchmark results and demos rate lower than
production systems in real organizations.
- 1 — No product does this; research-stage only.
- 3 — Products exist but with material error rates or narrow scope.
- 5 — Mature products perform this reliably in production at scale.

### cost_ratio (1–5)
Compare the AI cost per task-equivalent (inference + integration + oversight)
to the loaded human wage for the same output.
- 1 — AI is more expensive than the human, all-in.
- 3 — Roughly comparable cost.
- 5 — AI is at least an order of magnitude cheaper.

### barriers (1–5) — NOTE: higher = MORE protected
Strength of adoption barriers: licensing/authorization requirements, liability
and error-cost asymmetry, regulatory coverage of the automation itself,
human-contact requirement, organizational friction.
- 1 — No meaningful barriers; nothing prevents substitution.
- 3 — Some friction (oversight requirements, customer preference for humans).
- 5 — Hard barriers: a licensed human must legally perform or sign off on the task.

### adoption_velocity (1–5)
How fast and deep are the sectors where this task occurs actually adopting AI
(agents in production, measured displacement), per public adoption data?
- 1 — Laggard sectors (small firms, physical, low digitization).
- 3 — Middling adoption, pilots common, production rare.
- 5 — Fast, deep adoption (information, finance, professional services patterns).

### augmentation (1–5)
Independently of replacement: does AI **assist** a human doing this task,
raising their productivity? High augmentation and low automatability can
coexist (assistive drafting for a task requiring human judgment).
- 1 — AI offers no meaningful assistance.
- 3 — Useful assistance on parts of the task.
- 5 — AI transforms productivity on this task while the human stays in the loop.

## Output

Return ONLY a JSON object of this shape (no prose outside JSON):

```json
{
  "automatability": { "rating": 1, "rationale": "..." },
  "feasibility": { "rating": 1, "rationale": "..." },
  "cost_ratio": { "rating": 1, "rationale": "..." },
  "barriers": { "rating": 1, "rationale": "..." },
  "adoption_velocity": { "rating": 1, "rationale": "..." },
  "augmentation": { "rating": 1, "rationale": "..." }
}
```