The correct option is The first is prompt engineering and the second is prompt tuning and prompt tuning typically reduces ongoing enablement effort while requiring greater upfront investment.

Teaching staff to write clearer and more targeted prompts is prompt engineering. This improves outputs by guiding the model with better instructions and examples while leaving model weights unchanged. Adopting a tool that adapts the underlying model to the organization’s voice and tasks aligns with prompt tuning on Vertex AI. This creates a tuned variant that teams can reuse across workflows which lowers the ongoing enablement burden. However it requires upfront data collection, configuration, evaluation, and a tuning run which is why the initial investment is higher.

The first is few shot prompting and the second is full model retraining on Vertex AI is incorrect because few shot prompting is only one pattern within prompt engineering rather than the full scope of the first approach. It is also incorrect to describe the second approach as full model retraining because Vertex AI uses lightweight tuning methods such as prompt tuning or adapters for foundation models rather than retraining the entire base model.

Both are prompt engineering and they require the same level of investment is incorrect because the second approach involves tuning a model variant which demands upfront effort for data preparation and evaluation while prompt engineering alone typically has lower initial cost.

The first is prompt tuning and the second is prompt engineering and they demand the same technical depth is incorrect because the approaches are reversed and the effort is not the same. Prompt engineering emphasizes user guidance and iteration while tuning introduces additional setup and evaluation that increase the initial technical depth.

Full AWS Practitioner Certification Question