AWS AI Business Strategist (AIB-C01) — Practice Questions and Mock Exam

The tenets give a centre of excellence a defensible remit and a way to measure itself. Research decides which areas to explore and which practices to build. Evangelize spreads knowledge through events, publications and training. Apply turns that into an end-to-end delivery roadmap, and Lead proves the capability through pilots and first wins. Mentor is the one aimed at capability transfer beyond the team itself, and Scale turns what worked into reusable patterns. A centre of excellence that only delivers projects has quietly dropped several of these. Exam tip: match the tenet to the verb in the stem, since onboarding and coaching others map to Mentor while publishing and awareness map to Evangelize.

Business goal identification is the phase where a programme is still allowed to conclude that it does not need a model at all. The lens asks for a clear statement of the problem and of the business value expected, and only then for a judgement on whether machine learning is the right way to get there. Alternatives are compared on the accuracy they would deliver, their cost and how well they scale, which is a business comparison rather than a technical one. Skipping this step is how organisations end up maintaining a model where a rule, a report or a process change would have done the job. It is also the moment to check that enough relevant, high-quality training data actually exists. Exam tip: an option about algorithms, instances or retraining is answering a later phase than the one the question asks about.

The customisation ladder exists because each rung costs more and commits more. Prompt engineering changes nothing permanent, can be revised in an afternoon, and resolves a surprising share of quality complaints, which is why it is the recommended first step. Retrieval augmentation is the right rung when the gap is knowledge rather than behaviour, and it fits volatile material particularly well because updating the source updates the answers without touching the model. Fine-tuning changes the weights and therefore produces an artefact that has to be retrained whenever the underlying content moves, which makes it a poor fit for a handbook revised weekly and a reasonable one for consistent tone, format or task behaviour. Naming the gap correctly is what selects the rung.

Foundation model endpoints are a new kind of front door, and the data that flows back through them is often more sensitive than the request that opened them. Identity is the layer that holds regardless of where the model is hosted, which is why the lens frames least privilege as protection that applies to any hosting arrangement rather than a substitute for network controls. Sessions and permission boundaries add limits in time and in scope, so a role that legitimately needs access does not carry it indefinitely or beyond its purpose. The lens also puts these boundaries in the organisation's AI policy document rather than in individual project documentation, so they can be reviewed as part of a regular access review and applied consistently across every workload rather than being renegotiated each time.