# AI as Decision Support

> AI supports The Leapers' assessment methodology inside defined competencies, behavioral indicators and human review — it does not own the decision.

Source: https://www.theleapers.co/en/methodology/ai-as-decision-support

- AI is most useful when its boundaries are clear.
- At The Leapers, AI works inside a methodology defined by competencies, behavioral indicators, exercise context and human quality control.
- It structures data, surfaces patterns and produces decision-support outputs that human experts can use.

AI as Decision Support

TWO AI LAYERS

## Simulation Layer

- May play a role-play persona
- May act as a controlled information interface in Fact Finding
- Supports structured interaction
- Creates consistent digital participant experiences

It does not freely improvise beyond the intended assessment environment.

## Analysis & Decision-Support Layer

- Processes transcripts and written responses
- Structures evidence
- Matches evidence against defined indicators
- Supports BARS/BOS-based preliminary analysis
- Identifies patterns
- Supports report drafting
- Generates development-oriented insight

Every step operates within predefined competencies, indicators and scoring logic.

## What AI does — and does not do

- Structure evidence
- Process transcripts
- Analyze defined behavioral content
- Identify predefined patterns
- Support cross-exercise evidence integration
- Support report drafting
- Support controlled role-play / fact-finding experiences

- Make the final hiring decision
- Independently decide promotion
- Decide who is “high potential”
- Freely infer personality
- Evaluate undefined behaviour
- Make high-impact conclusions from emotion alone
- Close final high-impact reports without human review

TRUST ARCHITECTURE

Six layers of trust, plus contextual adaptation

## Methodological Trust

- Assessment target defined
- Competencies selected
- Indicators defined
- Exercise and scoring logic aligned

## Structural Trust

- AI works inside Behavioral Competency Architecture
- Not free-form interpretation

## Technical Guardrails

- AI operates within predefined scope
- Simulation access is controlled
- Undefined evidence is not freely interpreted

## Explainability

- Outputs can be connected to competencies and behaviour
- Reports produce insight, not unexplained scores

## Human Quality Control

- Final expert review remains
- High-impact decisions are not fully automated

## Proportionate Data Use

- Only relevant data is used
- Focus stays on evidence needed for the assessment

## Contextual layer — organization-specific adaptation

The methodological core remains stable, but the following can be customized:

- Competencies
- Indicators
- Role context
- Scenario
- Reporting emphasis

## Data transparency

### Primary data

#### Text

Written answers and transcripts.

#### Speech

Speech can be converted to text and analyzed within assessment context.

### Supporting signals where appropriate

Paralinguistic / tone signals may support interpretation.

They must not independently determine an assessment result.

### Currently not an active assessment data source

Video and facial-expression analysis are not active assessment data sources in the current methodology.

## Emotion and tone are supporting signals only

Emotion, tone or paralinguistic signals alone do not produce an assessment result.

Primary evaluation remains based on content, behavioral patterns, competency indicators and exercise methodology.

## Human oversight, by delivery model

### Live model

A human assessor may observe the interaction directly.

### Digital model

A live assessor may not be present during participant interaction. But:

- Methodology is human-designed
- Behavioral boundaries are predefined
- Scoring / observation logic is predefined
- Final human review remains

Digital does not mean “fully autonomous”.

## Live, digital and hybrid are delivery models — not quality tiers

The methodological core remains consistent across all three:

- Behavioral architecture
- Structured evidence
- BARS/BOS where appropriate
- AI guardrails
- Human review

## Operationalizing digital assessment with TAP

Parts of The Leapers' assessment methodology can be operationalized through TAP in digital and hybrid assessment environments.

- Digital exercises
- Evidence collection
- AI-supported analysis
- Structured reporting
- Assessment workflows

[Explore TAP](https://usetap.ai)

TAP is not The Leapers' internal platform, nor a sub-brand — and this methodology is not limited to working only with TAP.

## Does AI make the final decision?

No. AI is positioned as decision support, and high-impact outputs remain subject to human review.

## Can AI be biased?

Any human or technology-supported assessment system can contain bias risk if poorly designed. The Leapers manages this risk through behavioral indicators, structured methodology, controlled AI use and human review.

## What data does AI use?

Text, transcripts and relevant supporting paralinguistic signals, within the defined assessment context.

## Do you use facial analysis?

No, it is not an active data source in the current methodology.

## What happens if the human assessor and AI produce different results?

Human expert review and evidence reconciliation are required before any final output.

## Can digital assessment be trustworthy without a human assessor present?

Yes, because the methodology is human-designed and a final human review remains, even when a live assessor is not present during the interaction.

## Trust starts with knowing where AI stops.

[Explore the Explainable Assessment System™](https://www.theleapers.co/en/methodology/the-leapers-explainable-assessment-system)

[Explore TAP](https://usetap.ai)
