The last lesson covers the capabilities most Claude users never discover: artifacts for generating live code and documents, data analysis from raw CSV data, and building real workflows that connect Claude to the tools you already use. Plus an honest take on when Claude wins vs. when to use something else.
Artifacts are Claude's way of generating content that exists as a standalone object — not just text in the conversation. When you ask Claude to write code, build an interactive tool, or create a document, it can put the output in an Artifact panel that you can view, copy, edit, and iterate on separately from the conversation.
Artifacts are available on claude.ai (not via API by default). They appear in a side panel and support:
# Build an interactive tool
"Build an HTML calculator that estimates the ROI
of implementing AI automation in a business process.
Inputs: hours saved per week, hourly rate, implementation
cost, time horizon (months). Show result in real-time
as user types."
# Build a data visualization
"Create an HTML bar chart showing this monthly
revenue data:
Jan: 42000, Feb: 38000, Mar: 51000, Apr: 48000,
May: 63000, Jun: 71000
Make it clean, no libraries, pure HTML/CSS/JS."
# Build a template
"Create a professional project status report template
in markdown. Include sections for: executive summary,
key metrics, milestones (with traffic lights), risks,
next actions, and decisions needed."
Claude can analyze raw data you paste directly into the conversation. No Python required, no Jupyter notebook setup, no pandas imports. Just paste your data and describe what you want to know.
This works best with CSV data under ~500 rows. For larger datasets, use the Claude API with Python.
Here's a CSV of our sales data for Q1:
date,rep,region,product,amount,closed
2024-01-03,Sarah,West,Pro,12500,true
2024-01-05,Marcus,East,Starter,3200,true
2024-01-07,Sarah,West,Enterprise,48000,true
[...paste your data...]
Analyze this data and tell me:
1. Top 3 reps by total revenue
2. Which product generates the most revenue per deal?
3. Average deal size by region
4. Are there any deals that look like outliers?
Flag anything unusual.
5. If Q2 follows the same trend, what would you
project for total Q2 revenue?
Show me the calculations behind each answer.
import anthropic
import pandas as pd
# Load your data
df = pd.read_csv('sales_data.csv')
# Convert to string for Claude
data_str = df.to_csv(index=False)
client = anthropic.Anthropic()
message = client.messages.create(
model="claude-opus-4-5",
max_tokens=2048,
messages=[
{
"role": "user",
"content": f"""Analyze this sales data CSV:
{data_str}
Find:
1. Top reps by revenue
2. Best performing product category
3. Monthly trend (improving/declining?)
4. Any anomalies worth investigating
Be specific with numbers."""
}
]
)
print(message.content[0].text)
Claude doesn't need to be a standalone chat tool. Here's how people actually integrate it into real workflows:
Connect Claude to Gmail, Slack, Notion, Airtable, and hundreds of other tools without code. Common automations: auto-summarize emails, generate first drafts from form submissions, classify support tickets.
For anything that needs to run automatically or at scale. See the Python snippet above. The API is $3-15 per million tokens depending on the model — for most business use cases, this is pennies per task.
Claude.ai has direct integrations with Google Drive, GitHub, and Jira. Connect them in Settings → Integrations. Then you can say "summarize the last 5 pull requests in my repo" or "find all open Jira tickets assigned to me."
No AI is best at everything. Here's a genuine comparison based on real use:
| Use Case | Best Choice | Why |
|---|---|---|
| Long document analysis | Claude | Better quality on full documents, more accurate extraction |
| Writing & editing | Claude | Better tone control, more nuanced understanding of voice |
| Coding (general) | Claude / GPT-4o (tie) | Both strong; Claude slightly better at explaining its code |
| Image generation | ChatGPT (DALL-E) | Claude doesn't generate images |
| Web browsing / current events | ChatGPT or Gemini | Claude's web access is limited; others are more robust |
| Google Workspace integration | Gemini | Native Docs/Sheets/Gmail integration is excellent |
| Very long context (>200K tokens) | Gemini 1.5 Pro | 1M token window, though quality varies |
| Honest "I don't know" responses | Claude | Less likely to hallucinate confidently |
| Plugin ecosystem | ChatGPT | GPT Store has hundreds of specialized plugins |
| API cost at scale | Roughly comparable | All three are in the same range; check current pricing |
The practical answer for most knowledge workers: use Claude as your primary tool for reading, writing, and analysis. Keep ChatGPT for image generation and web browsing. If you're a heavy Google Workspace user, try Gemini for Workspace tasks.
Map out one repetitive work process end-to-end and build Claude into it. This is the capstone exercise.
You've now been through 5 days of practical Claude training. Here's the real challenge: identify the single task in your work that takes the most time and produces the most frustration. Build a Claude workflow for it this week — custom instructions, knowledge base, prompt templates, the whole thing. One week from now, measure the difference. That's what this course was for.
You've covered the full spectrum — basics, documents, writing, Projects, and advanced workflows. The next step is deeper: building AI systems, not just using them.
Join the 3-Day Bootcamp →