1.3. The AWS AI Landscape at a Business Level
💡 First Principle: Cloud providers package AI in layers that trade control for convenience. The higher the layer, the faster you get value and the less you can customize; the lower the layer, the more you can differentiate and the more you must build, staff, and govern. Choosing a layer is a business decision about speed, cost, differentiation, and risk, and the responsibility for data, access, and appropriate use stays with you at every layer.
You do not need hands-on experience with any AWS service for this exam, but you do need to recognize what each named service is for, how it is priced, and which business situation calls for it. The exam guide names a short list: Amazon Bedrock, Amazon SageMaker AI, Amazon Quick, the AWS Cloud Adoption Framework (AWS CAF), the AWS shared responsibility model, and AWS cost tools (AWS Pricing Calculator, AWS Cost Explorer, AWS Marketplace). That short list is a strong hint: the exam tests strategic positioning of a few services, not a catalog.
A useful mental model is the difference between renting a furnished apartment, renting an empty apartment, and building a house. The furnished apartment (a ready-made application) is ready today but you live with the landlord's furniture. The empty apartment (a managed platform) gives you the structure and utilities and lets you furnish it your way. Building a house (custom models) gives you exactly what you want, at the highest cost and longest timeline, and you are responsible for every pipe.
⚠️ Common Misconception: "Using a managed AWS AI service transfers responsibility for data, access, and appropriate use to AWS," and "Bedrock and SageMaker AI are interchangeable." Managed services shift operational burden, not accountability for how you use them. And the two services sit at different layers: Bedrock gives access to existing foundation models; SageMaker AI is for building custom ML.
