Tau
AI agents and automation · 11 lessons

What are AI agents, and how can they work for you?

Hand AI jobs that take several steps, like comparing prices or sending you a weekly summary, and set limits so you stay in control. For people who already chat with AI and want it to do more.

Who it’s for: Confident AI users, professionals and small-business owners.

A chatbot answers your question. An automation repeats the same fixed steps every time, like saving email attachments to a folder. An AI agent sits in between. You give it a goal, and it chooses the steps to reach it. It might open websites, compare prices and put the results in a table.

This course shows you how to hand over jobs that take several steps to AI while you stay in control. You'll find the agent mode in the assistant you already use and give it a clear goal. Then you'll check its work against the real websites. You'll set up a scheduled task that gathers your week's meetings, deadlines and news into one Monday message. You'll also connect an assistant to your calendar or email, give it only the access it needs and learn how to disconnect it.

Safety runs through every lesson. You'll add approval steps, so nothing is sent, bought or deleted without your say-so. You'll test an agent on past cases before it touches anything real. You'll also learn when not to hand over a task at all. It's written for people who already chat with AI and want it to do more, including professionals and small-business owners.

By the end you’ll be able to…

  • Tell a chatbot, an automation and an agent apart with examples
  • Delegate a multi-step web task to an agent and review its work
  • Connect an assistant to your email or calendar with minimal access
  • Build a scheduled workflow that summarises new messages daily

Short on time? Start with the essentials

These 4 lessons cover the most useful ideas in this course. Each one takes about a minute.

Essential 1

What's an AI agent, exactly?

Classify tools as chatbots (answer), automations (repeat fixed steps) or agents (choose steps toward a goal).

Essential 2

Let AI compare prices for you

Give a browser agent a clear goal and output format, watch it work, and check its table against the sites.

Essential 3

Let AI see my calendar, safely

Connect an assistant to your calendar or email with the narrowest permissions, and know how to revoke them.

Essential 4

How do I keep an agent in check?

Configure approval steps, spending or action limits and activity logs before letting an agent act for you.

Every lesson in this course

This course has 11 video lessons of about a minute each, from the basics to expert habits. Each one ends with a question, three takeaways and its sources.

Foundations 2 lessons

  1. What's an AI agent, exactly?Essential

    Classify tools as chatbots (answer), automations (repeat fixed steps) or agents (choose steps toward a goal).

    “A chatbot answers. An automation repeats. An agent decides its own next step.”

  2. Does my AI app already have an agent?

    Locate the agent mode in ChatGPT, Gemini, Claude or Copilot and describe what it can and can't do.

    “ChatGPT Work, Gemini Tasks, Claude Cowork, Copilot Cowork: one idea, four names.”

Building skills 3 lessons

  1. Let AI compare prices for youEssential

    Give a browser agent a clear goal and output format, watch it work, and check its table against the sites.

    “Three websites, one table of prices and cancellation terms, while you make coffee.”

  2. Let AI see my calendar, safelyEssential

    Connect an assistant to your calendar or email with the narrowest permissions, and know how to revoke them.

    “Give an AI read-only access first. It can't send what it can't touch.”

  3. Get a Monday briefing, done for you

    Create a scheduled AI task that gathers your week's meetings, deadlines and news into one message.

    “Every Monday at 8am: your meetings, deadlines and weather, already summarised.”

Advanced 4 lessons

  1. How does AI plug into my other apps?

    Explain the Model Context Protocol as a standard plug between AI apps and tools, and judge whether a connector is trustworthy.

    “MCP is like USB-C for AI: one plug that lets assistants use your tools.”

  2. Get AI to sort new forms for you

    Split a repetitive task into trigger, AI step and action, and give the AI step a fixed output format so the next step can use it (new form response → summary → team chat).

    “Every new form response, summarised and posted to your team chat. No code.”

  3. How do I keep an agent in check?Essential

    Configure approval steps, spending or action limits and activity logs before letting an agent act for you.

    “Never let an agent send, buy or delete without a tap from you. Here's the setup.”

  4. When shouldn't I hand a task to AI?

    Decide against delegation when stakes, irreversibility, unclear goals or a cheap fixed automation make an agent a poor fit.

    “If a mistake costs more than doing it yourself, don't delegate it.”

Expert 2 lessons

  1. How do I plan a job for an agent?

    Choose between a fixed workflow (chain, router, parallel) and an open-ended agent, and sketch the steps and checks.

    “Anthropic's advice for agents: start with the simplest workflow that works.”

  2. Is my agent ready for real work?

    Run an agent on ten past cases, score its results and set a pass bar before it touches live data.

    “Run your agent on 10 past cases before you let it near a live inbox.”

Apps we recommend

These are the tools this course uses, and what each one is best for. We have no partnership with any of them.

Zapier

Best for: No-code workflows across thousands of apps

Connect forms, email, Slack and sheets, with AI steps, without writing code.

Freemium

n8n

Best for: Visual workflows you can self-host

Open, visual workflow builder with AI agent nodes; self-host it to keep data in house.

Freemium

Claude

Best for: Delegating multi-step tasks to an agent

Cowork runs multi-step tasks across your files, browser and connected apps, with scheduled runs.

Freemium
Hands-on project

Hand one real weekly chore to an agent, safely

Pick a recurring task from your work or home life, such as a Monday briefing, form summaries or inbox triage. Decide between a fixed workflow and an agent, connect the least access it needs, add approvals, test it on past cases, then run it for two weeks with tools you are allowed to use.

What you’ll make: A working workflow or agent, a one-page design (steps, access, approvals) and a test log of 10 past cases.

Check your own work

  • The task is written as a goal, its inputs, an output format and a done-condition
  • Every connected app has the narrowest access, listed with how to revoke it
  • Nothing is sent, bought or deleted without your approval
  • Tested on 10 past cases against a pass bar set before go-live, with the score recorded
  • Uses only tools your employer or household allows, with no confidential data in unapproved tools

Further reading and viewing

We’ve chosen these articles, research papers, videos and guides from other trusted sources.

Article · Anthropic · 20 min

Building effective agents

The go-to explainer on workflows vs agents and the simple patterns that work.

Guide · Model Context Protocol · 10 min

What is the Model Context Protocol (MCP)?

Official intro to the open standard that connects AI apps to tools and data.

Video · Sequoia Capital (Andrew Ng) · 13 min

What's next for AI agentic workflows

Andrew Ng on reflection, tool use, planning and multi-agent patterns.

Course · Hugging Face · 10 hours

AI Agents Course

Free hands-on course on building agents with tools, frameworks and evaluations.

Guide · OWASP · 20 min

OWASP GenAI Security Project

Community-standard risks and mitigations for LLM apps and agents.

Questions about this course

What is the difference between an AI agent and a chatbot?

A chatbot answers what you ask, one message at a time. An agent works towards a goal you give it, choosing and carrying out the steps itself. A chatbot can tell you how to compare flights. An agent can open the sites, collect the prices and hand you a table. That extra freedom is why agents need clear limits.

Are AI agents safe to use?

They can be, with sensible limits. Give an agent the narrowest access it needs, such as reading your calendar rather than managing your whole account. Ask it to get your approval before it sends, buys or deletes anything. Check its activity log and its results. Be aware, too, that hidden instructions on a web page can try to trick an agent.

What tasks can I automate with AI?

Good candidates are tasks you repeat often, with a clear result you can check. Think of a weekly summary of your meetings and deadlines, sorting new form responses, or comparing prices across a few shops. Tasks that are high-stakes, hard to undo or vaguely defined are poor fits. For those, it's usually better to keep a person in charge.

Do I need to code to build an AI automation?

No. Tools such as Zapier and n8n let you connect everyday apps with a trigger, an AI step and an action, without writing code. Many assistants, including ChatGPT, Gemini, Claude and Copilot, also have agent features built in. The most important skill is clearly describing the goal, the result you want and how you'll check it.

Who is this course for?

It's for people who already use AI chat and want it to take on bigger jobs. Lessons run from foundations to expert, so the first ones explain the basic ideas before you build anything. If you're brand new to AI, a beginner course is a better first step. Each lesson takes about a minute, and your first five lessons in Tau are free.

Get early access to this course

Tau opens soon. Join the waitlist and we’ll email you as soon as you can start this course.