Day 5 of 5 — Final Lesson
⏱ ~75 minutes
How to Use Claude — Day 5

Advanced Claude: Artifacts, Analysis, and Automation

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.

Using Artifacts

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:

Artifact Prompts That Work
# 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."
💡
Iterate on artifacts. Once Claude generates an artifact, you can say "make the chart taller," "add a dark mode toggle," or "add a fourth input field." Claude updates the artifact without starting over. This is pair programming at its most useful.

Data Analysis with Claude

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.

Prompt — CSV Data Analysis
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.

For larger datasets

Python — Claude API for Data Analysis
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)

Connecting Claude to Your Workflow

Claude doesn't need to be a standalone chat tool. Here's how people actually integrate it into real workflows:

Via Zapier / Make

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.

Via the Claude API

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.

Via Claude.ai integrations

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."

ℹ️
Start with the highest-friction task. Look at your week and find the thing that takes the most time and requires the least judgment. That's your first automation. Common wins: summarizing meeting notes, first-draft responses to common emails, formatting data between systems.

Claude vs. ChatGPT vs. Gemini: The Honest Take

No AI is best at everything. Here's a genuine comparison based on real use:

Use CaseBest ChoiceWhy
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.

📝 Day 5 Exercise
Build a Complete Workflow for a Real Work Process

Map out one repetitive work process end-to-end and build Claude into it. This is the capstone exercise.

  1. Pick a process you do at least weekly: meeting follow-ups, customer emails, report generation, data review, content creation — anything with clear inputs and outputs.
  2. Write out the steps of the process as it exists today (without AI).
  3. For each step, decide: can Claude do this? Can Claude assist? Or should I keep doing it manually?
  4. Build the Claude part: write the prompts, create the Project, upload the knowledge files it needs.
  5. Run the process once with Claude integrated. Time it. Compare to your old time.
  6. Write 3 sentences on what worked, what didn't, and what you'd change next time.

Day 5 Summary

  • Artifacts turn Claude into an interactive tool builder — live HTML, code, documents you can iterate on in real-time.
  • Paste CSV data directly for quick analysis. For larger datasets or automation, use the API with Python.
  • Integrate Claude via Zapier, Make, or direct API — start with the highest-friction manual task in your week.
  • No AI is best at everything. Claude leads on document analysis, writing, and honest responses. Use ChatGPT for images and web; Gemini for Google Workspace.
Final Challenge

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.

Course Complete.

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 →