Quick Start

Pick your starting point

An AI model is the "brain" of your agent. Pick whichever feels most familiar — they all work for the steps that follow.

Give it a real task

Don't just say "tell me about AI." Give it a task you actually do at work. The key is specificity — the more context you provide, the more useful the response.

Here are some concrete examples you can use right now:

Customer communications:

"Here are my last 5 customer emails [paste them]. Draft a professional response to each one that acknowledges their concern, provides a clear next step, and maintains a friendly tone."

Meeting follow-up:

"Here's the transcript from today's team meeting [paste it]. Summarize the key decisions made, list all action items with the responsible person's name, and flag any unresolved questions."

Data and spreadsheets:

"I have a spreadsheet of invoices with columns: Invoice Number, Client, Date, Amount, and Status. Write an Excel formula that flags any invoice where Status is 'Open' and the Date is more than 30 days ago."

Research and synthesis:

"Here are three articles about [topic]. Identify the main points of agreement, any areas where the authors disagree, and summarize the overall picture in 3 bullet points for a non-expert audience."

Notice how the model completes the work — not just answers a question. That's the shift from chatbot to agent.

The pattern in each example: provide the raw material, describe the task clearly, specify the format you want back.

Level up: use a tool-enabled agent

Claude and other models can be given tools — the ability to browse the web, run code, read files, or call APIs. This is where agents become genuinely autonomous.

Try Claude.ai Projects (free tier available):

1. Create a new project
2. Upload a document (a PDF, a spreadsheet, meeting notes — anything)
3. Ask it to analyze, extract, or act on the content

For example, upload your company's pricing sheet and ask: "A customer wants X units of product A and Y units of product B with a 15% enterprise discount. Calculate the total and draft a quote email."

The agent reads the document and uses that knowledge to complete the task — you didn't have to copy-paste the pricing information manually.

What tool use unlocks:

• Agents that can look things up rather than relying only on what they were trained on
• Agents that can run calculations and verify their own work
• Agents that can interact with other software on your behalf

Automate a workflow

Ready to go further? Pick one of the Recipes and follow it step by step. Each recipe shows exactly how to set up an agent for a specific business task — no coding required for most of them.

Good starting points based on your role:

• Sales: Sales Prospecting — research leads and draft outreach automatically
• Operations: Invoice Automation — flag, categorize, and route invoices
• Marketing: Content Repurposing — turn one piece of content into many formats
• HR: Employee Onboarding — guide new hires through your process

Build your own

If you want to build custom agents and deploy them — either for yourself or for your team — head to the Tools section. We'll walk you through the full stack: picking a model, choosing a harness or platform, wiring up tools, and deploying something your team can actually use.