1.2.2. What the Strategist Decides and What Builders Decide
💡 First Principle: Good collaboration depends on knowing whose decision a decision is. When business leaders make technical calls, they choose by brand name or hype; when technical teams make business calls, they optimize accuracy while ignoring cost, risk, and adoption. The exam rewards answers that keep each decision with the people best equipped to make it.
The exam guide lists tasks that are explicitly out of scope for the AI business strategist. They are a useful map of the boundary:
| The strategist decides | Technical teams decide (out of scope for AIB-C01) |
|---|---|
| Which business problems are worth solving with AI | Which algorithm, framework, or architecture to use |
| Whether AI is appropriate at all, or whether rules or process change suffice | How to engineer features and prepare data |
| Build, buy, or partner, based on value, cost, time, capability, and regulation | Hyperparameter tuning and model optimization |
| KPIs, baselines, ROI method, and cost guardrails | Building and deploying ML pipelines and infrastructure |
| Acceptable risk, required human oversight, and governance structure | Implementing security and compliance protocols in systems |
| Readiness investments: data ownership, skills, culture, sponsorship | Configuring and administering AWS services |
| When to scale, pause, or terminate an initiative | Monitoring infrastructure, debugging pipelines, troubleshooting model performance |
| How to communicate change and redesign roles | Data cleaning, labeling, and annotation |
Two nuances matter. First, "out of scope" does not mean "not your concern." A strategist must require that monitoring exists, that access is restricted, and that data is labeled well. The strategist sets the requirement and holds people accountable; the technical team decides how to meet it. Second, the boundary shows up in answer choices. If a question asks what a program manager should do about a model whose accuracy is falling, "work with the data science team to establish monitoring thresholds and a retraining trigger tied to business KPIs" is in role; "adjust the learning rate and retrain" is not.
⚠️ Exam Trap: An answer can be technically correct and still wrong for this exam because it describes an engineering action. Ask yourself: "Would a product manager or consultant personally do this?" If not, look for the option that sets a requirement, makes a decision, or engages the right team.
Reflection Question: A vendor's proposal recommends a specific neural-network architecture. What questions should the strategist ask about the proposal, and which decision should be left to the technical evaluators?
