What to Read
This section covers the basic language of GenAI leadership: what generative AI can do,
where it fits in an organization, and how to separate augmentation from automation.
Key question: should AI assist a human, or should it run a complete workflow with review gates?
Augmentation vs Automation
| Pattern | Meaning | Good fit |
|---|
| Augmentation | AI assists a human, but the human makes the important judgment. | Writing, summarizing, analysis support, high-context decisions. |
| Automation | AI handles a task end to end under defined rules and review gates. | Repeatable workflows, routing, extraction, low-risk operational tasks. |
The Five GenAI Layers
- Infrastructure: compute, storage, networking, GPUs, TPUs, and runtime foundations.
- Models: Gemini, Gemma, Imagen, Veo, embeddings, and other foundation models.
- Platform: Vertex AI and managed services to build, tune, deploy, and evaluate.
- Agents: systems that reason, plan, call tools, and act toward goals.
- Applications: the user-facing assistant, automation, or product workflow.
Solution Selection Checklist
- Start from the business problem and measurable outcome.
- Decide whether the use case requires text, image, video, code, search, or action.
- Check whether the system needs current enterprise data or only general model knowledge.
- Define human review points before building the demo.
- Choose the simplest tool that meets the governance and reliability bar.
