What Claude Can Do

Welcome to AI Workflows Mastery. This course is designed for working professionals — specifically L&D specialists and Product Managers — who want to use Claude as a daily productivity tool, not a novelty.

By the end of this first lesson, you will have a clear mental model of what Claude can and cannot do, and when to reach for it instead of other tools.

Claude's Core Capabilities

Claude is a large language model built by Anthropic. Think of it as a reasoning engine that works with text, code, and structured data. Here is what it does well:

1. Text Analysis & Comprehension

• Summarize long documents (research papers, meeting transcripts, legal contracts)
• Extract specific information from unstructured text (pull action items from a 40-page report)
• Compare documents side by side (highlight differences between two policy versions)
• Classify content (sort customer feedback into categories, tag support tickets by urgency)

2. Content Creation & Writing

• Draft professional documents (emails, reports, proposals, training materials)
• Rewrite content for different audiences (turn a technical spec into an executive summary)
• Expand outlines into full text (turn bullet points into polished paragraphs)
• Edit for tone, clarity, grammar, and style consistency

3. Reasoning & Analysis

• Break down complex problems step by step
• Evaluate options with structured pros and cons
• Identify gaps, risks, and edge cases in plans
• Generate frameworks for decision-making

4. Code & Technical Work

• Write code in any major programming language
• Explain existing code in plain English
• Debug errors with suggested fixes
• Convert between formats (CSV to JSON, SQL to plain language)

5. Artifacts & Structured Output

• Create tables, charts, and formatted documents
• Generate slide outlines, spreadsheet formulas, and templates
• Produce structured data (JSON, XML, Markdown)
• Build interactive prototypes and simple web pages

Claude vs. Other LLMs

| Capability | Claude | ChatGPT | Gemini | |-----------|--------|---------|--------| | Long document analysis | Excellent (200K+ tokens) | Good | Good | | Following complex instructions | Excellent | Good | Good | | Nuanced writing tone | Excellent | Good | Moderate | | Code generation | Excellent | Excellent | Good | | Image understanding | Yes | Yes | Yes | | Web browsing | No (by design) | Yes | Yes | | Data privacy controls | Strong (enterprise) | Varies | Varies |

The key difference: Claude is designed to be helpful, harmless, and honest. It will tell you when it is uncertain rather than confidently generating incorrect information.

When to Use Claude vs. Search

Use Claude when:

• You need to create something (drafts, plans, outlines)
• You need to transform content (rewrite, summarize, restructure)
• You need to analyze text you provide (meeting notes, survey results)
• You need to reason through a problem (evaluate options, find gaps)

Use search when:

• You need current facts (today's stock price, breaking news)
• You need to verify specific claims (exact regulatory requirements)
• You need real-time data (flight status, weather)

What Claude Cannot Do

• No internet access in the standard interface
• Knowledge cutoff date — does not know about very recent events
• No memory between conversations unless using Projects
• Can generate plausible-sounding but incorrect information
• Cannot execute code on your systems

Exercise: Capability Mapping

List three tasks from your last workweek that match Claude's capabilities:

1. A document you had to summarize or analyze
2. Content you had to create from scratch
3. A decision you had to reason through

Prompting Best Practices

The difference between a mediocre AI output and a genuinely useful one is almost always the prompt.

The CRAFT Framework

CRAFT is a repeatable structure for writing professional prompts:

C — Context

Tell Claude the background information it needs. Weak: "Write a training email." Strong: "Our company just adopted Asana. 200 employees need to transition from spreadsheets within 30 days. Most are non-technical."

R — Role

Tell Claude who it should be. Weak: "Help me write this." Strong: "You are a senior L&D specialist with 10 years of experience designing corporate training programs."

A — Action

State exactly what you want Claude to do. Use clear verbs: write, analyze, compare, list, draft, evaluate, summarize, create.

F — Format

Specify the output structure. Tables, bullet points, numbered lists, headers, word counts.

T — Tone

Define the voice and style. Professional, casual, encouraging, direct, empathetic.

Advanced Prompting Techniques

Chain-of-Thought

Ask Claude to reason step by step before giving a final answer. This dramatically improves accuracy on complex tasks.

Few-Shot Examples

Show Claude what good output looks like by providing 1-3 examples.

Structured Output Requests

Ask for specific formats that are easy to use in your workflow (JSON, tables, numbered lists).

Iterative Refinement

Start broad, then refine in 2-3 follow-up turns.

Common Prompting Mistakes

| Mistake | Problem | Fix | |---------|---------|-----| | Too vague | Generic output | Specify topic, audience, length, format | | No context | Claude guesses | Provide background in 2-3 sentences | | No role | Generic tone | Assign a relevant expert role | | No format | Wall of text | Request tables, bullets, headers | | Overloading | Confused output | Break into sequential prompts | | Not iterating | Accepting first output | Refine in 2-3 follow-ups |

Exercise: Rewrite Bad Prompts

Practice by rewriting these weak prompts using CRAFT:

1. "Write me a survey." → Apply all five CRAFT elements
2. "Summarize this document." → Add context about who needs the summary
3. "Help me plan a project." → Specify project type, timeline, team size, deliverable format