Choose and Set Up Your AI Tools

People who get the least from AI usually have four subscriptions and use none of them well. This lesson gets you to one assistant, configured properly, in thirty minutes.

---

1. The one-tool rule

For your first two weeks, pick one assistant and use it for everything — even when another would be better for a particular task.

That sounds backwards. It isn't. You learn what a tool is genuinely good at by hitting its limits, and you only hit its limits by using it for everything. Sampling four tools teaches you their marketing.

You can add a second tool later, once you know exactly what gap it fills.

2. Choosing

Don't choose from a benchmark or a comparison article — rankings change every few months, and the differences at the top are smaller than the marketing suggests.

Choose from what you mostly do:

If you mostly...	Prioritise
Write emails, documents, long-form	Prose quality and how well it follows instructions
Research things and need sources	Live web search with real, clickable citations
Work in Google or Microsoft files	Whichever connects to the suite you already use
Handle sensitive or client data	Data controls and the right paid tier
Work with very long documents	A large context window

For which products currently lead each category, see the AI Tool Radar in the resources below. It's the one page in this course that tracks names and prices, and it carries a date so you know how fresh it is.

The Decision Tree resource walks through this in more detail if you're torn.

3. Free or paid?

Start free. Upgrade when you hit the free tier's limits repeatedly in a week, doing work that matters — not before.

A rough rule: if a subscription costs about what an hour of your time is worth, it pays for itself the first month it saves you an hour.

4. Set it up properly

This is the part people skip, and it's what makes the difference between a tool and a toy.

Security, first:

• Turn on two-factor authentication.
• Turn off "use my conversations to improve the model" — it's usually under Settings → Data Controls or Privacy. Next lesson covers this in depth.

Custom instructions, second. Most assistants let you set instructions that apply to every conversation. Paste this and edit it to describe you:

About me: [ROLE, INDUSTRY, WHO YOU WORK WITH]
I am usually trying to: [TYPICAL GOALS]
My audience is usually: [WHO READS WHAT YOU PRODUCE]

How to respond:
- Get to the point. No preamble, no summary of my question.
- If my question is ambiguous, ask one clarifying question first.
- Say when you are unsure rather than guessing confidently.
- Plain English, no corporate filler.
- For anything factual, tell me what you would check.

You'll replace this with your full Digital Twin in lesson 4. This is the starter version.

Reach, third. Put the assistant somewhere you'll open it without thinking: phone home screen, pinned browser tab, a keyboard shortcut. If it takes thirty seconds to get to, you'll stop using it.

5. First test

Ask it to explain something in your own field — something you know well.

You'll immediately see its failure mode, because you'll spot the errors an outsider wouldn't. That calibration — knowing how confidently wrong it can be about something you understand — is the most useful thing you can learn in your first week.

---

⚠️ Common mistakes

• Choosing by benchmark. Choose by what you actually do.
• Subscribing to several at once. One for two weeks, then decide.
• Skipping custom instructions. Every conversation starts generic without them.
• Leaving training switched on. Takes ten seconds to turn off.
• Testing it only on things you don't know. You can't see its mistakes there.

---

What's next: twenty minutes of account security that prevents the mistakes people regret.