Domain 5 - Product and Model Selection (12%)¶
Overview¶
This domain tests whether you can match a business need to the right Claude product, plan, and model. Scenarios describe a person, a task, and constraints (cost, speed, quality, collaboration, governance), and ask what they should use. No configuration detail is required - that is Domain 6 - just sound selection judgment.
The Claude Product Family¶
Claude.ai chat¶
The general-purpose assistant in the browser, desktop, and mobile apps. Handles drafting, analysis, Q&A, document upload, image understanding, web search, and computation via the analysis tool. The default surface for one-off tasks.
Projects¶
Persistent workspaces inside Claude.ai. A Project carries:
- Custom instructions applied to every chat in the Project
- Knowledge files available to every chat in the Project
- On Team/Enterprise plans, sharing with teammates
Select a Project whenever the scenario mentions recurring work, a consistent workstream, repeated context-pasting, or a team needing the same setup.
Artifacts¶
The dedicated side-panel where Claude creates and iterates on substantial outputs: documents, reports, code snippets, diagrams, and small interactive pages. Select Artifacts when the scenario centers on producing and refining a deliverable, or something the user wants to view, edit iteratively, or share as a standalone piece.
Claude in Slack and other integrations¶
Claude embedded where work already happens: Slack (summarize threads, draft replies, answer questions in-channel), browser extension, and desktop-app conveniences. Select an integration when the scenario's friction is "the information and the conversation live in another tool and people copy-paste".
Connectors¶
Connections from Claude to live sources - drives, documents, and business tools - so Claude works from current data rather than uploads. Select connectors when the scenario stresses "always current" information or repeated re-uploading of changing files. Governance note: organizations control which connectors are enabled (Domain 3/6 overlap).
Claude Code¶
Anthropic's agentic command-line tool for software engineering: it reads codebases, edits files, runs commands. For the CCAO-F you need concept-level awareness only: it targets developers, and it is the answer when a scenario is about engineering teams automating coding work - and the wrong answer for a business analyst's document workflow.
The API and developer platform¶
Building Claude into products and automated pipelines is developer territory (the Developer Foundations cert). For this exam, recognize the boundary: when a scenario needs fully automated, high-volume, system-to-system processing, that is an API/engineering conversation to bring to a technical team - not a chat workflow.
Product Selection Cheat Table¶
| Scenario signal | Best fit |
|---|---|
| One-off question or draft | Claude.ai chat |
| Same workflow every week, same background docs | Project |
| Team should share one setup and context | Shared Project (Team/Enterprise) |
| Building and refining a document or mini-app | Artifacts |
| Work and discussion live in Slack | Claude in Slack |
| Needs live files/systems, uploads keep going stale | Connectors |
| Engineering team automating code tasks | Claude Code |
| Automated pipeline, no human in the chat | API (refer to developers) |
The Model Family¶
Claude models come in three tiers. Exact version names change; the tier logic is what the exam tests.
| Tier | Character | Best for |
|---|---|---|
| Opus | Most capable, deepest reasoning, slowest, highest cost | Complex analysis, long multi-step reasoning, hardest writing tasks where quality dominates |
| Sonnet | Balanced capability, speed, and cost | The default for most everyday business work |
| Haiku | Fastest and cheapest, lighter capability | High-volume simple tasks: classification, short summaries, quick extraction |
📖 Models Overview - the current model family, capabilities, and comparisons
Selection heuristics¶
- Default to the balanced tier. Most tasks do not need the flagship, and the balanced model is faster and cheaper.
- Step up when quality dominates - dense analysis, high-stakes deliverables, tasks the balanced model handles poorly after good prompting.
- Step down when volume and speed dominate - hundreds of short, similar, simple items.
- Prompt quality beats model choice for most quality problems. "Switch to a bigger model" is rarely the first fix (Domain 7 overlap): if the prompt lacks context, Opus will produce a better-written wrong answer.
- Extended thinking - Claude's option to reason longer before answering. Turn it on for genuinely hard problems; it costs time, so it is not a default for quick tasks.
Exam trap¶
"Always use the most capable model" is a distractor. So is "the cheap model for everything". The correct answers match the tier to the described task, and often mention trying a better prompt before a bigger model.
Plans¶
| Plan | Audience | Selection signals |
|---|---|---|
| Free | Individuals trying Claude | Light, occasional use; tight usage limits |
| Pro | Individual power users | Daily individual use, higher limits, full features for one person |
| Team | Small-to-mid teams | Shared billing, shared Projects and collaboration, commercial data terms |
| Enterprise | Larger organizations | Everything in Team plus admin controls, SSO-style access management, enhanced security and governance capabilities |
Selection logic the exam uses:
- Individual experimenting → Free or Pro.
- A team collaborating on client work → Team at minimum; the data terms and shared Projects are the point.
- An organization with compliance requirements, central user management, or audit needs → Enterprise.
- The governance rule from Domain 3 shows up here: client-confidential work on a personal Free/Pro account is a wrong answer even though it is technically possible.
Putting It Together: Selection Scenarios¶
"A consultant answers similar RFP questions every week from a library of past responses." Project (recurring workflow + standing knowledge files) on a Team/Enterprise plan (client data), balanced model. Artifacts for assembling each response document.
"A support lead wants 300 short customer emails per day categorized and given draft replies, no human typing prompts." That is an automated pipeline: refer to the engineering team for an API solution, likely on the fast/cheap model tier. A chat product is the wrong shape.
"A partner needs a one-time deep analysis of a 100-page market study for a board presentation." Claude.ai chat with the document uploaded; the most capable model tier (possibly with extended thinking) since quality dominates and volume is one; output drafted as an Artifact; every figure verified before the board sees it (Domain 1).
"A project team living in Slack keeps copy-pasting thread summaries into Claude." Claude in Slack.
Key Takeaways¶
- Match the surface to the work pattern: chat for one-offs, Projects for recurring work, Artifacts for deliverables, integrations for in-tool work, connectors for live data.
- Claude Code is for developers; the API is for automated pipelines - know the boundary and hand off.
- Model tiers: balanced by default, up for quality-dominant tasks, down for volume-dominant simple tasks.
- Better prompting usually beats a bigger model as the first fix.
- Plans are a governance choice as much as a feature choice: business data on business plans.
📖 Claude.ai - current plans and features 📖 Claude Help Center - product documentation for each feature