# Understanding AI Vulnerability

Cubit provides multi-dimensional intelligence about how AI will impact work. This page explains what each score means and how to interpret them.

## The Core Insight

**AI doesn't automate jobs-it automates tasks.**

A "Financial Analyst" performs dozens of distinct activities: gathering data (highly automatable), building models (partially automatable), and managing client relationships (protected). Treating the job as a single unit obscures what's actually happening.

Cubit analyzes work at the **task level**, then aggregates to jobs with proper importance weighting.

## The Three Pillars

Every occupation is evaluated on three independent pillars:

### 1. Structural Exposure

*"Can AI technically access and execute this work?"*

This pillar measures the "attack surface" available for automation:

| Component                 | What It Measures                                | High Score Means                                |
| ------------------------- | ----------------------------------------------- | ----------------------------------------------- |
| **Procedural Intensity**  | How rule-based and repeatable is the work?      | Scripted, predictable, follows decision trees   |
| **Digital Accessibility** | Can software observe inputs and act on outputs? | Entirely within digital systems, API-accessible |

**Structural Exposure = (Procedural + Digital) / 2**

Work that is both procedural AND digital provides natural entry points for AI systems.

### 2. Human Necessity

*"Even if AI could do this, is human presence essential?"*

This pillar identifies work protected by biological or social constraints:

| Component               | What It Measures                                      | High Score Means                                       |
| ----------------------- | ----------------------------------------------------- | ------------------------------------------------------ |
| **Physical Embodiment** | Does it require motor control, dexterity, navigation? | Hands-on, unstructured environments, fine manipulation |
| **Relational Depth**    | Does it require empathy, trust, persuasion?           | Deep interpersonal engagement, emotional labor         |

**Human Necessity = (Physical + Relational) / 2**

Patients want human doctors. Students want human teachers. Clients want human advisors. Some work is protected by human preference, not technical limitations.

### 3. Demonstrated Capability

*"Can current AI systems actually perform the required skills?"*

Unlike the structural pillars, this one is grounded in **empirical evidence**:

1. We identify the skills, abilities, and knowledge required for each task
2. We map those requirements to standardized AI benchmarks (HELM)
3. We measure how well state-of-the-art models actually perform
4. We aggregate performance back to tasks and jobs

This prevents both:

* **False alarms**: Flagging work as "at risk" when AI can't actually do it yet
* **Missed risks**: Assuming work is "safe" when AI already excels at the underlying skills

## Composite Scores

The three pillars combine into actionable metrics:

### Automation Susceptibility Score (0-100)

```
Automation Susceptibility = AI Exposure Potential x 100
```

Where AI Exposure Potential blends Structural Exposure with Demonstrated Capability.

| Range  | Interpretation                                      |
| ------ | --------------------------------------------------- |
| 0-35   | Low exposure - limited AI entry points              |
| 35-50  | Moderate exposure - some tasks vulnerable           |
| 50-65  | Elevated exposure - significant automation pressure |
| 65-100 | High exposure - most tasks accessible to AI         |

### Human-Centric Resilience Score (0-100)

```
Resilience = (Human Necessity + Dispositional Resilience) / 2 x 100
```

Dispositional Resilience comes from O\*NET Work Styles-personality factors like adaptability, interpersonal orientation, and stress tolerance that make humans valuable regardless of task content.

| Range  | Interpretation                                 |
| ------ | ---------------------------------------------- |
| 0-40   | Low resilience - limited human necessity       |
| 40-55  | Moderate resilience - some protective factors  |
| 55-70  | Good resilience - meaningful human requirement |
| 70-100 | High resilience - strong human essentiality    |

### Balanced Impact Score (-100 to +100)

```
Balanced Impact = Resilience Score - Susceptibility Score
```

This is the bottom line:

| Range         | Interpretation  | Implication                               |
| ------------- | --------------- | ----------------------------------------- |
| +20 or higher | Net positive    | Strong position; human factors dominate   |
| +5 to +20     | Slight positive | Good position; monitor for shifts         |
| -5 to +5      | Tension zone    | Augmentation likely; role will evolve     |
| -5 to -20     | Slight negative | Pressure building; prepare for transition |
| -20 or lower  | Net negative    | High displacement risk                    |

## Example: Three Jobs Compared

| Job                     | Automation Susceptibility | Resilience | Balanced Impact |
| ----------------------- | ------------------------- | ---------- | --------------- |
| **Registered Nurses**   | 47                        | 70         | +23             |
| **Software Developers** | 48                        | 55         | +7              |
| **Bookkeeping Clerks**  | 49                        | 49         | 0               |

**Interpretation:**

* **Nurses**: Despite moderate AI exposure, strong human necessity (physical care, patient relationships) provides protection
* **Software Developers**: In the augmentation zone-AI assists but human judgment remains central
* **Bookkeeping Clerks**: Tension zone-role likely to transform significantly

## Quantile Scores

In addition to raw scores, Cubit provides **quantile rankings** (0-1):

```python
job = client.get_job("15-1252.00")
print(job["scores"]["automation_susceptibility_quantile"])  # 0.58
```

This means 58% of occupations have lower automation susceptibility. Quantiles help you:

* Rank occupations relative to each other
* Identify outliers
* Build percentile-based visualizations

## Important Caveats

{% hint style="info" %}
**Scores reflect positioning, not prediction.**

A high automation susceptibility score doesn't mean the job will disappear tomorrow. It means the job is structurally accessible to AI and the required skills are increasingly within AI's reach. Actual displacement depends on:

* Adoption rates and investment decisions
* Regulatory environment
* Organizational change capacity
* Economic factors beyond task-level analysis
  {% endhint %}

{% hint style="info" %}
**Scores are relative to today's AI.**

Demonstrated Capability scores reflect current model performance. As AI improves, these scores will shift. Cubit is versioned and periodically refreshed.
{% endhint %}

## Next Steps

* [The Four Dimensions](/core-concepts/four-dimensions.md) - Details into the scoring components
* [Quadrant Classification](/core-concepts/quadrants.md) - How jobs are categorized
* [Data Model](/core-concepts/data-model.md) - Jobs, tasks, requirements, and benchmarks


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.maidenlabs.tools/core-concepts/understanding-scores.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.