Insights Navigator - HypersightAI

DS Data Structure

How structured and machine-readable is the input data?

4

Fully structured — clean databases, APIs, standardised fields. Machine can read it directly with no pre-processing.

3

Mostly structured — minor gaps or mixed formats. Light normalisation needed before processing.

2

Semi-structured — documents, emails, forms with variable layouts. Requires parsing or NLP extraction.

1

Unstructured — free text, images, audio, video. Significant pre-processing required; high noise.

DL Decision Logic

How rule-based vs. judgment-driven is the decision?

4

Pure rules — deterministic if/then logic. Outcomes are fully predictable from inputs alone. Zero ambiguity.

3

Mostly rules — a few edge cases need human review, but the mainstream path is fully rule-driven.

2

Mixed — guidelines exist but significant contextual judgment is applied case-by-case by experienced staff.

1

Pure judgment — every decision relies on intuition, relationship context, or expertise that can't be codified.

VF Volume / Frequency

How many times does this process run?

4

Thousands per day — high-frequency repetitive work. Automation ROI is immediate and compounding.

3

Hundreds per week — regular batch or queue-based work. Strong automation case once volume compounds.

2

Tens per month — moderate frequency. Automation helps but build cost must be weighed against savings.

1

Rare / one-off — happens infrequently or never repeats exactly. Automation cost exceeds benefit.

ER Exception Rate

How often do edge cases require human judgment?

4

Rarely (<5%) — edge cases are uncommon and easy to route to a human exception queue without disrupting flow.

3

Occasionally (5–20%) — exceptions exist but are identifiable in advance. Can be handled by routing rules.

2

Frequently (20–50%) — a large share of cases hit edge conditions. Human involvement is structurally required.

1

Almost always (>50%) — nearly every instance requires human judgement. Automation would create more problems than it solves.

RC Regulatory Compliance

How much regulatory risk does automation carry?

4

Low risk — no regulatory constraint on automation. Errors are recoverable; no compliance audit trail required.

3

Some constraints — regulations exist but don't prohibit automation. Logging and explainability requirements are manageable.

2

Moderate risk — regulated domain with explainability or human-in-loop requirements. Automation possible but needs governance.

1

High risk — strong regulatory mandate for human oversight. Automated decisions may be prohibited or carry major liability.

SD Strategic Differentiation

Is human involvement a competitive advantage here?

Override rule: SD = 1 caps the classification at Augmented AI regardless of total score. SD is a ceiling, not just a contributor.

4

Commodity task — industry-standard, undifferentiated work. Automating frees humans for higher-value activities. No override.

3

Low differentiation — mostly standard. Automating has little competitive downside; incremental human touch adds modest value.

2

Moderate differentiation — human creativity or relationship quality adds real value. Automation should augment, not replace.

1

Core differentiator — human judgment, creativity, or relationships are the product. Automating destroys competitive advantage. → Caps at Augmented AI