Daily use for staff

25 minutes. For everyone. The right workspace, the right model, and clean text when it leaves ChatGPT.

---

This lesson is for everyone on the team, not just the owner. If you're the owner, sit with one staff member and do it together. You'll both learn something.

There's a reason to do this now, even for people who've used ChatGPT for a year. On October 7, 2026, OpenAI began rolling out GPT-6 as the default model in Chat for Business workspaces. It answers differently, and its answers can include more than text. Habits that worked last month need a quick check.

Step 1: Make sure you're in the company workspace (2 minutes)

If you also have a personal ChatGPT account, you may have two workspaces under one login. Work belongs in the company one.

On the web, your workspaces are in the profile menu.
On the phone app, they're in the sidebar.

Pick the company workspace before you start. Look for your company's name (the owner set it in Lesson 1). Make it a habit, like checking which email account you're sending from.

If you see no company workspace, you haven't accepted your invitation yet. Find the email and accept it with the same address it was sent to.

Step 2: Know the model menu (5 minutes)

Open a new chat and look at the model menu. Here's what OpenAI's pages say you'll find in a Business workspace:

GPT-6 is the default as it rolls out. It replaced the option called Latest. If you had Latest selected, you were moved to GPT-6 automatically. If you had picked GPT-5.6 Sol on purpose, you stay on it until you change it.
A Thinking slider sets how hard it thinks: Instant, Medium, High, Extra High and Pro.
Pro uses a separate, stronger model (GPT-6 Pro). OpenAI says Business Standard seats get 15 Pro messages a month and Premium seats get 50 a week, shared with GPT-5.6 Sol Pro. Save it for the rare hard problem.

If GPT-6 isn't in your menu yet, OpenAI says it may not have reached your workspace. That's fine. Everything in this lesson still applies.

A simple rule for the team:

Instant for quick rewrites, short emails and simple questions.
Medium or High when it needs to work something out: comparing quotes, checking a schedule, reading a long document.
Pro only when Medium or High wasn't good enough, and the answer really matters.

One more setting to know. OpenAI's Business page says ChatGPT can add reasoning on its own while Instant is selected, and this can use your reasoning allowance. You control it in Settings > General, under Higher intelligence. Most people can leave it alone.

Step 3: Expect answers that aren't just text (5 minutes)

GPT-6 brings a feature OpenAI calls Intelligent UI. OpenAI says answers can now include graphics, tappable buttons, forms, charts and interactive tools, like a side-by-side comparison or a calculator, right in the conversation. ChatGPT picks the layout based on the question. Sometimes it's plain text. Sometimes it isn't.

That's useful when you're working something out. It's a problem when the answer is meant to leave ChatGPT. A chart, a button or a form lives in the chat. An email, a text message or a work order needs plain words.

So, two habits:

When the answer is going somewhere else, ask for plain text. Add a line like this to your request:

> Give me plain text only, ready to paste into an email. No tables, buttons, charts or special formatting.

Test before you trust it. The first time you paste a GPT-6 answer into your email, invoicing or scheduling software, send it to yourself first. Check that nothing went missing and nothing looks odd. Do this once per tool, and you'll know.

If you'd rather see fewer visual elements everywhere, OpenAI says you can reduce them in Settings on ChatGPT on the web, using the Layout and visuals option. OpenAI notes that some visual elements may still appear. You can also add a standing preference in your custom instructions, such as "keep answers short and in plain text unless I ask for a table."

One more note from OpenAI: these features don't work in the older ChatGPT desktop apps for Mac and Windows. If something looks broken there, open ChatGPT on the web instead.

Step 4: Write requests that get good answers (5 minutes)

Nothing about GPT-6 changes the basics. A good request has four parts:

Who it's for. "This is a reply to a homeowner."
What you want. "Draft a short email confirming their appointment."
The facts. "Tuesday, between 8 and 10. Technician is Luis. Furnace tune-up."
The format. "Plain text, under 80 words, friendly, no sign-off."

Here's Dana at Comfort Zone Heating & Cooling, our fictional HVAC company, writing an appointment confirmation:

> I'm the office manager at a heating and cooling company. Draft a short email to a homeowner confirming a furnace tune-up on Tuesday between 8 and 10 a.m. with our technician Luis. Ask them to clear a path to the furnace. Plain text only, under 80 words, friendly, no sign-off.

Every fact in that request came from Dana. ChatGPT didn't have to guess any of them.

Step 5: Keep a person in charge (3 minutes)

The rules on the Team ChatGPT Card:

ChatGPT drafts. A person decides. Anything sent to a customer gets read, fixed and sent by a person.
Check every fact against your own records. Prices, dates, times, names, part numbers, amounts. If a fact in the draft didn't come from you, treat it as wrong until you've checked it.
Money stays with people. Quotes, invoices, refunds, discounts and payroll can be drafted with help. A person checks the numbers and approves them.
Keep out what doesn't belong. No passwords, door codes or card numbers. Only the customer details the task needs.
Use a temporary chat for one-off questions you don't need to keep.

Step 6: Start your retest log (5 minutes)

Every team has a few requests they send again and again. Write down your three most-used ones on the Prompt Retest Log now. Copy them exactly, or as close as you can.

Then run each one once on GPT-6, in the company workspace. Note in a sentence how the answer compares to what you're used to: better, same, worse, or "different format." You'll come back to these in Lesson 7.

Do this now

Switch to the company workspace on your computer and your phone.
Find the model menu and the Thinking slider.
Ask for one real draft, with "plain text only" in the request, and paste it into your email as a test to yourself.
Fill in your three prompts on the Prompt Retest Log.

Check it worked

You can find the company workspace on every device you use for work.
You know which Thinking level to use for everyday tasks, and that Pro is limited.
A pasted draft arrived in your email looking right.
Your three most-used prompts are written down, with a first note on GPT-6.
You can say the person-in-charge rule from memory.

---

Next: stop retyping the same instructions. Save them once and share them with the team.

→ Lesson 4: Projects, custom GPTs and sharing

Previous: Settings that protect you · Next: Projects, custom GPTs and sharing