Main Concept
AWS AI/ML managed services are pre-built, fully managed services that provide AI/ML capabilities through simple API calls — no model training, no infrastructure management, and no ML expertise required. You consume the capability as a service and pay only for what you use.
Why Managed Services Instead of Building Your Own
Key Idea
Lower barrier to entry → no ML expertise needed to add AI capabilities to an application.
Speed to market → integrate AI features in days instead of months.
Cost-effectiveness → no upfront investment in data collection, training infrastructure, or ML talent.
Accessibility → any developer can consume AI capabilities via API calls.
Efficiency → AWS handles infrastructure, scaling, and model maintenance.
Ability to meet business objectives → focus on the business problem, not the underlying technology.
Analogy: Electricity vs building your own power plant
You could build your own power plant to get electricity — but why would you? AWS managed services are like the power grid: you plug in and consume the capability immediately, paying only for what you use. Building your own model is the power plant — expensive, complex, and only justified when the standard service does not meet your specific needs.
The Services — By Capability
Language & Text
| Service | What it does |
|---|---|
| Amazon Comprehend | Natural language processing — sentiment analysis, entity detection, language detection, topic modeling |
| Amazon Translate | Neural machine translation between languages |
| Amazon Textract | Extracts text and structured data (forms, tables) from documents and images |
| Amazon Lex | Build conversational interfaces — chatbots and voice bots (the technology behind Alexa) |
| Amazon Kendra | Intelligent enterprise search powered by ML — searches across documents and data sources |
Speech
| Service | What it does |
|---|---|
| Amazon Transcribe | Automatic speech recognition — converts audio/speech to text |
| Amazon Polly | Text-to-speech — converts text to natural-sounding audio |
Vision
| Service | What it does |
|---|---|
| Amazon Rekognition | Computer vision — image and video analysis, object detection, facial recognition, content moderation |
Personalization & Recommendations
| Service | What it does |
|---|---|
| Amazon Personalize | Real-time personalization and recommendation engine — the same technology used by Amazon.com |
Fraud & Risk
| Service | What it does |
|---|---|
| Amazon Fraud Detector | Detects potentially fraudulent activity using ML — online fraud, account takeover, payment fraud |
Generative AI & Foundation Models
| Service | What it does |
|---|---|
| Amazon Bedrock | Fully managed access to foundation models from multiple providers (Anthropic, Meta, Mistral, Amazon Nova, etc.) via API |
| Amazon Q | Generative AI assistant for business — Q Business (enterprise knowledge), Q Developer (coding) |
| Amazon SageMaker JumpStart | Pre-built ML solutions and foundation models ready to deploy |
Human Review
| Service | What it does |
|---|---|
| Amazon Augmented AI (A2I) | Adds human review workflows into ML predictions — for cases where model confidence is low |
ML Platform
| Service | What it does |
|---|---|
| Amazon SageMaker | End-to-end ML platform — build, train, tune, deploy, and monitor custom ML models |
Managed Service vs Custom Model — When to Use Which
Key Idea
Use a managed service → when the capability is general-purpose and your use case is standard (speech-to-text, translation, sentiment analysis, image recognition).
Use a custom model (SageMaker) → when your use case is highly specific, proprietary data gives you a competitive advantage, or no managed service covers your need.
Use Amazon Bedrock → when you need foundation model capabilities with flexibility to choose providers and customize via RAG or fine-tuning.
Decision examples
“Add subtitles to our videos” → Amazon Transcribe (managed, no model needed)
“Detect our proprietary product defects from factory photos” → custom model on SageMaker (too specific for Rekognition alone)
“Build a customer service chatbot” → Amazon Lex (managed) or Amazon Bedrock (if needs generative responses)
“Recommend products to users” → Amazon Personalize (managed recommendation engine)
Exam Quick Reference
Speech to text → Amazon Transcribe
Text to speech → Amazon Polly
Translation → Amazon Translate
NLP / sentiment → Amazon Comprehend
Document extraction → Amazon Textract
Enterprise search → Amazon Kendra
Chatbot / voice bot → Amazon Lex
Image / video analysis → Amazon Rekognition
Recommendations → Amazon Personalize
Fraud detection → Amazon Fraud Detector
Foundation models → Amazon Bedrock
Generative AI assistant → Amazon Q
Human review loop → Amazon Augmented AI (A2I)
Custom ML models → Amazon SageMaker
Exam Scope
This is a high-priority area — expect multiple scenario questions asking you to match a business use case to the correct AWS service, and questions asking you to justify the advantages of using managed services. Memorize the quick reference table above.
Exam Domain
- Domain 1, Task Statement 1.2: “Explain the capabilities of AWS managed AI/ML services (for example, SageMaker, Amazon Transcribe, Amazon Translate, Amazon Comprehend, Amazon Lex, Amazon Polly).”
- Domain 2, Task Statement 2.3: “Describe the advantages of using AWS generative AI services to build applications (for example, accessibility, lower barrier to entry, efficiency, cost-effectiveness, speed to market, ability to meet business objectives).”
- Domain 2, Task Statement 2.3: “Understand the benefits of AWS infrastructure for generative AI applications (for example, security, compliance, responsibility, safety).”