Domain 1 of 4 · Chapter 2 of 3

Identify and select appropriate AI solutions

Unlock the complete study guide.

Bundled into the existing AWS Certified AI Business Strategist premium course — no separate purchase.

14-day refund guarantee on one-time purchases and your first subscription charge.

Included in this chapter:

  • Start with the process, then choose the solution type
  • Rule-based automation or AI: the write-it-down test
  • What makes a solution an AI agent
  • Human oversight: blocking review versus notification
  • When a multi-agent workflow is worth its overhead
  • Orchestration patterns: who delegates and integrates
  • Sequential or parallel: dependencies decide
  • Making agent collaboration reliable
  • Why AI solutions need ongoing monitoring
  • Responding to drift after it is detected
  • Classifying AI tools in a status registry
  • Finding and reducing shadow AI
  • Exam-pattern recognition for solution selection

Rule-based automation and AI solution types: who decides the next step

Decision factorRule-based automationAI model stepSingle AI agentMulti-agent workflow
Who decides the next stepRules written in advanceThe surrounding workflow; the model interprets, predicts, or generatesThe agent, after observing intermediate resultsA supervisor agent or peer agents coordinating specialists
Best fitComplete, stable rules such as thresholds and calculationsVaried wording or layouts inside a defined processGoal-directed work whose next step needs judgment about what is found that no complete rule can captureDistinct specialties or independent subtasks
PredictabilitySame input, same pathVariable output inside a fixed pathPath can differ between runsVariation can compound across handoffs
Main oversight needControl over rule changesAgreed quality thresholds and exception handlingBlocking review before consequential actionsHandoff contracts and end-to-end evaluation
ExampleRoute invoices above a fixed limit to a directorExtract fields from invoice PDFs in many layoutsResolve a return request by checking order recordsParallel security, cost, and compliance checks, then synthesis

Decision tree

First: map and simplifythe processCan you write this step'srule down completely?YesRule-basedautomationNoNext step needs judgmentno complete rule captures?NoAI model stepin a workflowYesSpecialist agents' benefitexceeds orchestrationoverhead?NoSingleAI agentYesMulti-agentworkflowAlways: monitor against a baseline

Cheat sheet

  • Complete and stable rules favor deterministic automation
  • A workflow can combine AI interpretation with deterministic business controls
  • Automation choice follows an understood business process rather than a desire to use AI
  • An agent pursues a goal rather than merely providing a conversational answer
  • Agents can revise a plan by observing the results of intermediate actions
  • Tool use connects an agent's reasoning to current information and business actions
  • Agent memory preserves working context without necessarily changing the underlying model
  • Human oversight can require approval before action or provide visibility without blocking
  • Multiple agents are justified by distinct responsibilities, not by agent count alone
  • Orchestration defines who delegates, coordinates, and integrates agent work
  • Dependencies determine whether agents should work sequentially or in parallel
  • Agent-to-agent communication must preserve the information needed for the next responsibility
  • Correct individual agents do not guarantee a correct collaborative workflow
  • Ongoing monitoring needs a reference point and business-relevant response thresholds
  • Input changes and changed business relationships are different forms of drift
  • Prediction quality requires comparison with observed outcomes, not just healthy inputs
  • Retraining timing should follow evidence of need rather than an arbitrary calendar alone
  • An updated model must demonstrate improvement before replacing the current one
  • Monitoring delivers value only when findings feed an owned improvement process
  • A shared tool registry makes approval status visible when employees need it
  • AI-tool approval must identify the uses and data that the approval actually permits
  • Shadow AI can appear inside software the organization already approved
  • Practical approved alternatives reduce the incentive to use unsanctioned AI
  • Approval records and actual-use visibility answer different shadow-AI questions

Unlock with Premium — includes all practice exams and the complete study guide.

References

  1. https://aws.amazon.com/blogs/machine-learning/best-practices-for-building-agentic-automations-with-amazon-quick-automate/
  2. https://docs.aws.amazon.com/aws-certification/latest/ai-business-strategist-01/ai-business-strategist-01-domain1.html
  3. https://docs.aws.amazon.com/aws-certification/latest/ai-business-strategist-01/ai-business-strategist-01-domain2.html
  4. https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf
  5. https://docs.aws.amazon.com/aws-certification/latest/ai-business-strategist-01/ai-business-strategist-01-domain3.html
  6. https://docs.aws.amazon.com/wellarchitected/latest/responsible-ai-lens/raiuc01-bp01.html
  7. https://docs.aws.eu/prescriptive-guidance/latest/agentic-ai-serverless/orchestration-models.html
  8. https://docs.aws.amazon.com/wellarchitected/latest/responsible-ai-lens/raiuc01-bp02.html
  9. https://aws.amazon.com/what-is/ai-agents/
  10. https://aws.amazon.com/blogs/machine-learning/strands-agents-sdk-a-technical-deep-dive-into-agent-architectures-and-observability/
  11. https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/what-is-bedrock-agentcore.html
  12. https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/memory.html
  13. https://docs.aws.amazon.com/pdfs/wellarchitected/latest/agentic-ai-lens/agentic-ai-lens.pdf
  14. https://docs.aws.amazon.com/wellarchitected/latest/agentic-ai-lens/agentsec04-bp02.html
  15. https://aws.amazon.com/marketplace/build-learn/ai-agent-learning-series/multi-agent-architectures
  16. https://docs.aws.amazon.com/prescriptive-guidance/latest/agentic-ai-patterns/multi-agent-collaboration.html
  17. https://docs.aws.amazon.com/wellarchitected/latest/agentic-ai-lens/agentops01-bp02.html
  18. https://docs.aws.amazon.com/wellarchitected/latest/machine-learning-lens/mlops01-bp03.html
  19. https://docs.aws.amazon.com/wellarchitected/latest/responsible-ai-lens/raimon02-bp01.html
  20. https://docs.aws.amazon.com/prescriptive-guidance/latest/gen-ai-lifecycle-operational-excellence/prod-monitoring-drift.html
  21. https://docs.aws.amazon.com/sagemaker/latest/dg/sms.html
  22. https://docs.aws.amazon.com/sagemaker/latest/dg/model-monitor-model-quality.html
  23. https://docs.aws.amazon.com/wellarchitected/latest/machine-learning-lens/mlsus06-bp02.html
  24. https://docs.aws.amazon.com/wellarchitected/latest/machine-learning-lens/mlrel04-bp02.html
  25. https://docs.aws.amazon.com/sagemaker/latest/dg/clarify-model-monitor-bias-drift.html
  26. https://aws.amazon.com/blogs/migration-and-modernization/govern-ai-adoption-before-it-governs-you/
  27. https://docs.aws.amazon.com/wellarchitected/latest/agentic-ai-lens/agentops03-bp04.html
  28. https://docs.aws.amazon.com/whitepapers/latest/navigating-security-landscape-genai/shadow-generative-ai.html