5. Implement Computer Vision Solutions (10-15%)

This domain shows the AI-103 shift most vividly. Where the predecessor exam taught a dedicated vision service for every task, AI-103 expects you to recognize that multimodal generative models now handle most visual understanding through prompts, reserving specialized vision services and custom training for narrower needs. The skill being tested is choosing the right tool for the visual job — and not over-reaching for a custom model when a multimodal prompt or a prebuilt capability already does it.