Four things worth knowing before you use AI at work

I show small business owners how to use AI every week, at our AI for Business workshops here in Shaftesbury. The same few questions come up every time, so I made a short video covering them, and here is the written version if you would rather read it.

None of this is complicated. It is four habits, and once you know them, you will get better answers, faster, for less cost.

1. Use the plus button for files and connectors, not just typing

Next to the message box, there is a plus sign. Click it and you will see two options.

"Add files" lets you upload something from your computer: a PDF, a spreadsheet, a photo, a Word document. The AI reads it properly, rather than you having to describe what is in it.

"Connectors" is different. Instead of uploading a file, you are linking a whole app, your Google Drive, your inbox, your calendar, so the AI can pull from it whenever you ask.

Simple rule: one file, upload it. Something you will come back to again and again, connect it once.

2. Pick the model and effort level to match the job

You will see a model name next to the send button, and usually a second setting called effort, with options like Low, Medium, High or Max.

The model is which version of the AI you are using. A lighter model is fast and cheap, fine for quick, simple things. A heavier model thinks harder and does better on genuinely difficult work, but it is slower and uses more of your allowance.

Effort is a separate dial. Low or Medium is for routine stuff, answering an email, tidying up a paragraph. High or Max is for when you need it to properly reason something through, planning a campaign or checking a contract.

Match the effort to the task. Do not use Max effort for a one-line reply, and do not use Low effort when you actually need it to think.

3. Know the difference between dictation and voice mode

There are two separate microphone features.

Dictation just turns your speech into typed text in the message box, so you can talk instead of type, then edit before you send. I use this constantly.

Voice mode is a full spoken conversation. You talk, it talks back. By default it listens continuously and replies when you pause. There is also a push-to-talk option, where you hold the button while you speak and let go when you are done, useful in a noisy office.

4. Use fewer tokens, whichever tool you are on

Every message you send and every reply you get is measured in tokens, small chunks of text. The more tokens a conversation uses, the more it costs and the slower it runs. This is not a Claude thing specifically. ChatGPT and Gemini work exactly the same way, they just name the dial differently.

Four habits that make the biggest difference, on any of these tools:

1. Start a new chat for a new topic. A conversation that runs on for weeks gets slower and pricier, because the AI re-reads the whole history every time you send a message.
2. Upload or link a file instead of pasting big chunks of text into the chat box. The file sits there once. Pasted text gets reprocessed every single message.
3. Turn the effort or thinking setting down for anything simple, and only turn it up when the task genuinely needs it.
4. Ask for what you actually need. A tight, specific question gets a tight, specific answer. A vague one invites a long one.

That is it

The plus button for files and connectors. The model and effort setting for matching power to the task. Voice input for when typing is a hassle. And keeping your chats short and focused to save tokens.

Get comfortable with those four, and you are already ahead of most people using this stuff.