5.2.3. Technology and Infrastructure Requirements
💡 First Principle: Infrastructure determines whether an AI idea can move from a laptop to a production process that thousands of people rely on. The strategist does not design the architecture, but must assess whether the foundations exist to deploy, integrate, secure, scale, and operate AI, and must fund the gaps before they become delays.
Foundational requirements a strategist should assess at a business level:
| Requirement | Business question | Why it matters |
|---|---|---|
| Cloud foundation | Do we have a secure, well-governed cloud environment (accounts, identity, networking, guardrails) where AI can be built and run? | Without it, every initiative waits for basic environments |
| Integration | Can AI connect to core business systems (CRM, ERP, ticketing) through APIs? | Value requires outputs to reach the systems and people who act on them |
| Data platform | Is there governed storage and processing for the data AI needs? | Supports data readiness and reuse (5.2.1, 5.2.2) |
| AI platform and model access | Do teams have approved access to foundation models and ML tooling (for example, Amazon Bedrock and Amazon SageMaker AI) with guardrails? | Avoids each team procuring tools independently and creating shadow AI |
| Security and identity | Can we enforce least privilege, encryption, logging, and data residency for AI workloads? | Required for governance and compliance (4.2.3) |
| Operations and monitoring | Can we monitor performance, cost, drift, and responsible AI metrics in production? | Required for production-grade AI (5.4.5) |
| Scalability and cost management | Can capacity grow with demand, and can costs be attributed and controlled? | Prevents success from becoming a budget crisis (3.2.5) |
Managed services shift much of the undifferentiated infrastructure burden to AWS, which is one reason organizations at early maturity often start with managed platforms such as Amazon Bedrock rather than building their own model-hosting infrastructure. But managed services still require a sound cloud foundation, integration, identity, and governance, which remain the customer's responsibility under the shared responsibility model.
A common gap is the prototype-to-production gap: teams can build impressive prototypes in sandbox environments but have no approved path to deploy into production with security, integration, and monitoring. Assessing and funding that path is often the highest-leverage infrastructure investment.
⚠️ Exam Trap: Infrastructure readiness is about the ability to deploy, integrate, secure, monitor, and scale AI in production, not just access to powerful compute or models. When a scenario describes prototypes that cannot reach production, look for answers that build the production path (integration, security, monitoring, operating standards) rather than more prototyping capacity.
Reflection Question: A company's data scientists build promising models on their laptops, but none have been integrated with the order management system. What infrastructure requirements would you assess, and why might a managed AI platform help but not fully solve the problem?
