The Creative Palette: Generation Use Cases on Bedrock
From stunning art to complex logic. Experience the real-world applications of Amazon Bedrock's multi-modal foundation models.
Generative AI in the Wild
In Module 3, we discussed "GenAI Use Cases" broadly. Now, we look at them through the lens of Amazon Bedrock. Bedrock provides specific "Model Families" that are leaders in certain modalities.
To pass the AWS Certified AI Practitioner exam, you must be able to recommend a model type (and sometimes a specific model name) for a business scenario.
1. Text Generation (The "Logical" Domain)
The Models: Anthropic Claude, Meta Llama, Amazon Titan Text, Mistral.
High-Impact Use Cases:
- Summarization: Turning long transcripts into bullet points.
- Copywriting: Generating thousands of unique product descriptions for an e-commerce site.
- Logic & Coding: Solving math problems or writing React.js components. (Claude 3.5 Sonnet is the leader here in 2026).
- Extraction: Taking a "messy" email and outputting a clean JSON object with the user's name and complaint.
2. Image Generation (The "Visual" Domain)
The Models: Stable Diffusion, Amazon Titan Image Generator.
High-Impact Use Cases:
- Advertising: Creating lifestyle photos of a product (e.g., "A watch resting on a mossy stone by a river") without a photo shoot.
- Variation: Changing the background of a photo or "Inpainting" (adding an object to an existing image).
- Stock Photography Replacement: Generating unique, high-quality images for articles and blogs.
3. Multimodal Generation (The "Seeing" Domain)
The Models: Anthropic Claude 3 (Haiku/Sonnet/Opus), Amazon Titan Multimodal Embeddings.
High-Impact Use Cases:
- Visual Q&A: Uploading a photo of a broken engine and asking the AI, "Where is the leak coming from?"
- Accessibility: Automatically generating "Alt-Text" descriptions for thousands of images for visually impaired users.
- Chart Analysis: Uploading a screenshot of a sales chart and asking the AI to "Calculate the percentage growth between Q1 and Q2."
4. Selection Matrix for the Exam
| Goal | Complexity | Recommended Bedrock Model |
|---|---|---|
| Simple Chatbot | Low | Claude 3.5 Haiku / Llama 8B |
| Complex Legal Analysis | High | Claude 3.5 Opus |
| Create 1,000 Ads | Medium | Stable Diffusion / Titan Image |
| Summarize 100-page PDF | Medium | Claude (Large Context Window) |
graph TD
A[Business Need] --> B{What is the Output?}
B -->|Text| C{Simple or Complex?}
C -->|Simple| D[Claude Haiku / Mistral]
C -->|Complex/Coding| E[Claude Opus / Llama 70B]
B -->|Image| F[Stable Diffusion / Titan Image]
B -->|Analysis of Image| G[Claude 3 Multi-modal]
5. Summary: Picking the Right Tool
On the exam, AWS will often provide two correct methods but ask for the "Most Cost-Effective."
- If you use Claude Opus (the most expensive model) to summarize a 2-paragraph email, you are "Over-Engineering."
- You should pick Claude Haiku or Mistral for simple, high-speed, low-cost tasks.
Exercise: The Architect's Choice
A retail company wants to analyze photos taken by customers of "items they found in other stores" and have the AI write a humorous 1-sentence caption for a TikTok challenge. They have 1 million photos a day and need to keep costs low. Which model family/type should they use?
- A. Stable Diffusion (Image Gen).
- B. Amazon Titan Text (Text Gen).
- C. A Multimodal model (like Claude 3 Haiku).
- D. Amazon Rekognition.
The Answer is C! You need the AI to "See" the photo (Multimodal) and "Generate" a caption (GenAI). Claude 3 Haiku is the "Fast/Cheap" version of the multimodal family, making it perfect for 1 million requests a day.
Knowledge Check
?Knowledge Check
Which Bedrock capability allows you to use your company's own documents (stored in S3) to provide relevant context for an AI's answer without re-training the model?
What's Next?
We know what to use. Now let's learn how to use it. In our next lesson, we break down the High-level integration patterns like RAG and Fine-Tuning.