INI-004
·Em revisãoAI-generated product descriptions
Diego Almeida · MarketingInternal productivityCriticality LowAutonomy: Suggests actions
80Overall
Automatic evaluation
Business Value75
AI Quality80
Security85
Privacy95
Opinion
Appropriate use of generative AI — human stays in the loop. Low risk. Recommended to proceed.
Aieval qualifies and recommends — the final decision is always human.
Gaps
1- Define brand style guide for the prompt
Risks
1- Generic text if prompt is poorly calibrated
Recommendations
2- Create a versioned prompt library
- A/B evaluation on the first 500 descriptions
Discovery
Answers collected during intake
- Problem
- Catalog with 50k SKUs without standardized descriptions.
- Objective
- Generate first-pass description for human review.
- Impacted users
- Content team (6 people)
- Vendor / model
- Anthropic Claude via API · Generative LLM with curated prompts
- Data used
- Product spec sheet
- Integrations
- Internal PIM
- KPIs
- SKUs described/day · Approval-without-edit rate
- Estimated cost
- $8,000
- Risk flags
- Uses PII: NoCustomer-facing: NoProduction access: NoExternal sharing: No