Last updated: March 2026 | Reading time: 7 min
So here's the thing: I used to spend my mornings doing the same repetitive work over and over. Answering customer emails. Drafting content briefs. Extracting data from invoices. It was soul-crushing, honestly. Then I discovered the Claude API, and everything changed. Now I've got five automations running that collectively save me 12+ hours every week. I'm not a real developer—I learned just enough Python to make this work—and you can too.
1. Head to console.anthropic.com
2. Sign up or log in
3. Find API Keys in the nav
4. Create a new key and stick it somewhere safe
(Honestly, this took me way too long to figure out because I kept looking in the wrong menu.)
`bash
pip install anthropic
`
That's it.
`python
import anthropic
client = anthropic.Anthropic(api_key="your-key-here")
message = client.messages.create(
model="claude-sonnet-4-6",
max_tokens=1024,
messages=[
{"role": "user", "content": "Summarize the key benefits of automation for small businesses in 3 bullet points."}
]
)
print(message.content[0].text)
`
Boom. You just made your first API call. Claude answered your question. Now let's actually build something useful instead of just playing around.
Time saved: 3 hours/week
API cost: ~$2/month
I was drowning in support emails last year. Not a ton—maybe 20-30 a day—but the repetition was killing me. So I built this. The script reads incoming emails, figures out what they're about, and drafts responses you can review and send.
`python
import anthropic
client = anthropic.Anthropic(api_key="your-key-here")
def draft_response(customer_email: str, context: str) -> str:
"""Generate a draft response to a customer email."""
message = client.messages.create(
model="claude-sonnet-4-6",
max_tokens=500,
system=f"""You are a customer service assistant for a small business.
Business context: {context}
Write a helpful, friendly response to this customer email.
Keep it under 150 words. Be direct and solve their problem.
Sign off with 'Best, [Team Name]'.""",
messages=[
{"role": "user", "content": f"Customer email:\n\n{customer_email}\n\nDraft a response."}
]
)
return message.content[0].text
email = "Hi, I purchased your course last week but haven't received the login details. Can you help?"
context = "We sell online courses. Login details are sent within 24 hours of purchase to the email used at checkout."
response = draft_response(email, context)
print(response)
`
How I actually use this: Every morning I run it on my support inbox. It drafts responses for like 80% of emails. I spend five minutes reviewing, maybe edit a sentence or two, hit send. What used to eat up 45 minutes of my day is now done by 9 AM. The best part? I wrote most of this while waiting for my coffee to cool down.
Time saved: 4 hours/week
API cost: ~$5/month
Before I write anything, I need to know what I'm actually writing about. Search intent. Angle. Structure. Used to take me forever to figure this out. Now Claude does it in 30 seconds.
`python
def generate_content_brief(keyword: str, audience: str) -> str:
"""Generate a comprehensive content brief for a given keyword."""
message = client.messages.create(
model="claude-sonnet-4-6",
max_tokens=2000,
messages=[
{"role": "user", "content": f"""Create a detailed content brief for an article targeting the keyword: "{keyword}"
Target audience: {audience}
Include:
1. Suggested title (under 60 characters, compelling)
2. Meta description (under 155 characters)
3. Search intent analysis (what is the reader trying to accomplish?)
4. Suggested H2 subheadings (6-8)
5. Key points to cover under each subheading
6. Related keywords to include naturally (10-15)
7. Suggested word count
8. Content angle that differentiates from existing articles
9. Potential internal/external linking opportunities
10. CTA suggestion"""}
]
)
return message.content[0].text
brief = generate_content_brief(
"best AI tools for small business",
"small business owners who are not technical"
)
print(brief)
`
Time saved: 2 hours/week
API cost: ~$1/month
Every invoice that came in used to require manual data entry. The vendor name, invoice number, line items, total amount due—I'd copy it all into a spreadsheet by hand. Super boring. This script pulls all that structured data out automatically.
`python
import json
def extract_invoice_data(invoice_text: str) -> dict:
"""Extract structured data from invoice text."""
message = client.messages.create(
model="claude-sonnet-4-6",
max_tokens=500,
messages=[
{"role": "user", "content": f"""Extract the following from this invoice and return as JSON:
Invoice text:
{invoice_text}
Return only valid JSON, no other text."""}
]
)
return json.loads(message.content[0].text)
`
Time saved: 2 hours/week
API cost: ~$2/month
Write one blog post. Now you need it on Twitter, LinkedIn, Instagram. Different tone for each platform. Different length. Different vibe. It's exhausting, and it usually meant that post got repurposed lazily (if at all). This automation turns one article into five polished social posts.
`python
def repurpose_for_social(article: str) -> dict:
"""Generate social media posts from an article."""
message = client.messages.create(
model="claude-sonnet-4-6",
max_tokens=1500,
messages=[
{"role": "user", "content": f"""Repurpose this article into social media posts.
Article:
{article[:3000]}
Generate:
1. Twitter/X thread (5-7 tweets, first tweet is a hook)
2. LinkedIn post (under 300 words, professional tone, personal angle)
3. Instagram caption (under 150 words, casual tone, 5 relevant hashtags)
Format each clearly with headers."""}
]
)
return {"social_content": message.content[0].text}
`
Time saved: 1 hour/week
API cost: ~$1/month
I used to sit down after every client call and spend 20 minutes typing up notes. Action items. Decisions made. Open questions. This takes the transcript and turns it into clean, organized notes.
`python
def process_meeting_notes(transcript: str) -> str:
"""Process meeting transcript into structured notes."""
message = client.messages.create(
model="claude-sonnet-4-6",
max_tokens=1000,
system="You are a meeting notes processor. Be concise and action-oriented.",
messages=[
{"role": "user", "content": f"""Process this meeting transcript into structured notes:
{transcript}
Format:
]
)
return message.content[0].text
`
1. Use Claude Sonnet for almost everything. It's way cheaper than Opus and honestly does 90% of business automation just fine. Save Opus for the complicated stuff.
2. Don't ask for more tokens than you need. If the task needs 200 tokens of output, don't request 4096. You pay for what you use.
3. Leverage prompt caching if you're repeating the same system prompt. If you're processing dozens of items with the same instructions, Anthropic's caching can cut your costs by up to 90%. It's wild.
4. Batch things up using the Batch API. You can get 50% off if you're willing to wait a few hours for results. Perfect for running automations overnight.
5. Actually check your usage. Go to console.anthropic.com/usage like once a week. Set spending limits so you don't get a surprise bill.
| Automation | How Often | Monthly Cost |
|-----------|----------|-------------|
| Email responses | 10-20/day | $1-3 |
| Content briefs | 5-10/week | $3-5 |
| Invoice extraction | 10-20/week | $0.50-1 |
| Social media | 5-10/week | $1-3 |
| Meeting notes | 5-10/week | $0.50-1 |
| Total | | $6-13/month |
For the cost of a bad coffee subscription, you're cutting 12+ hours out of your week. I genuinely can't think of a better ROI from any tool I've ever bought.
1. Sign up at console.anthropic.com
2. Throw $10 in credits (that'll last you weeks of testing)
3. Copy the email responder code above
4. Change the system prompt to match your actual business
5. Run it on five real emails and see what you get
Start with one automation. Get it working. Make sure the output is good. Then add the next one. Don't try to automate your entire operation on day one—you'll get overwhelmed.
Disclosure: This is for educational purposes. Prices and availability change. Always review AI-generated content before you actually use it for anything that matters.