How do you decide between cost vs quality in AI systems?

### Signal to interviewer

I can resolve cost-quality tradeoffs with an outcome-first framework that aligns economics to user impact.

### Clarify

I would clarify use-case criticality, failure tolerance, pricing model, and retention sensitivity to quality.

### Approach

Use an outcome-weighted tradeoff curve: estimate quality uplift impact on business outcomes and compare against incremental cost by segment.

### Metrics & instrumentation

Primary metric: cost per successful outcome by use-case segment. Secondary metrics: retention lift from quality gains, margin impact, and escalation rate. Guardrails: trust-critical failure growth and service-level deterioration.

### Tradeoffs

Higher quality routes improve outcomes but reduce margin headroom. Lower-cost routes improve efficiency but risk trust and repeat usage.

### Risks & mitigations

Risk: over-optimizing for average cases; mitigate with segment-level thresholds. Risk: hidden quality debt; mitigate with periodic deep evals. Risk: reactive cost cutting; mitigate with pre-agreed operating bands.

### Example

A contract review assistant uses premium models for high-risk clauses and efficient models for formatting and extraction tasks.

### 90-second version

Choose cost-quality balance by segment and outcome value. Optimize for cost per successful result, protect trust-critical workflows, and update operating points as economics change.