Stop asking ChatGPT every single time.

Three ways to stop retyping inside ChatGPT

• Custom instructions: set your role, industry, the documents you often write and phrases to avoid, and every prompt gets shorter. Write down the standards you keep when you do good work, like "Emails formal, lead with the conclusion" or "Show amounts with and without tax."
• Projects: create a project per client or deal and keep its files and chats together, so Chat and Work start from the same context.
• GPTs (custom GPTs): lock in the pattern for one task, such as "a GPT that turns meeting notes into our format." Public GPTs in the GPT Store work on the free plan, but building one requires Plus or above. GPTs shared only inside your workspace can be built on Team or Business and above (as of October 2026).

How to build a GPT in 4 steps

1. Open "Create a GPT." Building requires a paid plan, Plus or above; the free plan can only use GPTs (as of October 2026).
2. Set the instructions, knowledge files, conversation starters and capabilities. It's fine to have ChatGPT draft the instructions for you.
3. Choose who can access it. Only you, anyone with the link, or your organization (on Team or Business and above).
4. Use it. Examples: an advisor that has read your company documents, or a GPT that summarizes news from a URL.

Even so, you open the GPT, paste the files or URL, and ask, every time. SIMY takes over from picking up the work in meetings and chats.

Wait-for-input prompts: a trick that works on the free plan

Start with "Turn the bullet points I send next into prose using these rules. Reply 'OK' if you understand," then send only the data.

In long chats it gradually forgets the first instruction, so start a new chat and set the rules again when it drifts.

Even so, you open the chat, re-paste the rules and carry the data every time. SIMY learns your way of working from conversations and moves forward without instructions each time.

What's still left

A GPT is a pattern for one task. You decide when it starts, paste the transcript, carry the result to where it belongs, and open the same GPT again at the next meeting. In long chats the first instruction stops working and you may have to start over.

Scroll the table sideways →

How to choose: if you call up one pattern yourself, GPTs are enough. If the same kind of meeting happens every week and the work after it is the same every time, that's where SIMY comes in.

1. Check the training setting On by default
   Settings → Data Controls → turn off "Improve the model for everyone." It is on by default on the Free, Go, Plus and Pro personal plans, so check it before entering client names or amounts. Business and Enterprise data is not used for training by default.
2. Put your standards in custom instructions. Your role, industry, the documents you write and phrases to avoid. Write the standards you keep when you do good work, like "Emails formal, lead with the conclusion."
3. Separate deals with projects. Keep files and chats together per client or deal, so Chat and Work start from the same context.
4. Decide the scope of memory. With memory on, ChatGPT answers with past chats in mind, but you may not want it to remember client names or amounts. Decide up front whether to limit it to projects or turn it off.
5. Decide where answers go and what starts the next task. Don't stop at copying the answer. Decide whether it lives in Notion, Google Drive or Slack. Having a set place is the entry point that connects your work to SIMY.

A pattern for asking: goal, audience, structure, constraints

When asking for an email or similar, write these four first and you'll fix less.

Example: "Goal: reschedule a meeting. Audience: a director at a client. Structure: apology → three time options → request for reply. Constraints: under 80 words, formal."

Even so, you write the pattern and ask every time. SIMY picks up "who gets what" from meetings and chats and takes it as far as a draft, your way.

US dollar prices are as of October 2026, excluding tax. Prices vary by region, so check the ChatGPT pricing page for yours. Ads and GPTs details are from public information as of October 2026.

Scroll the table sideways →

Which plan: Go shows ads, just like Free. If you want ChatGPT to do real work, Plus is the minimum: Work and Codex, and no ads. For long Work sessions choose Pro; for teams that want chats kept out of training from day one, choose Business. Pro 200 stopped accepting new sign-ups on September 10, 2026 (existing subscriptions continue). Business is for teams of 2 to 200 employees.

For the latest details, see the ChatGPT pricing page. SIMY is priced separately from ChatGPT; see our pricing page.

Since July 2026, the ChatGPT desktop app brings Chat, Work and Codex together in one place. Their strengths differ clearly, so switch modes before you ask.

Scroll the table sideways →

How to use ChatGPT Work

ChatGPT Work, released on July 9, 2026, is an agent for long, complex tasks. It researches and analyzes across your connected apps and files, then finishes documents, spreadsheets, presentations, reports and Sites. ChatGPT working through tasks on its own began with the earlier "agent mode," and Work now carries that role.

1. Switch to Work. In the desktop app use the top-left menu; on web and mobile use the dropdown. The desktop app runs on macOS and Windows.
2. Choose where it runs. Turn on "Work locally" to let it use files and apps on your computer. Run in the cloud to keep going after you close your laptop.
3. Connect plugins. Slack, Google Drive, email, calendar and more connect via plugins. The old App Directory has been replaced by the Plugin Directory.
4. Include where to stop in your instructions. Set approval points up front, like "Stop before entering the quote amount and ask me to confirm," and you'll redo less.
5. Review the plan and approve. Work shows a plan before it starts. Answer its questions along the way and approve important actions.

Scheduled Tasks and usage limits

Work can schedule one-time runs, repeating runs such as every Monday at 9:00, and change monitoring. Since August 2026, MCP events from supported plugins can trigger automations too. Work and Codex share usage limits, with a cap every 5 hours. Pro gets 5x Plus, and Business premium seats have no 5-hour cap. Work itself has no extra fee.

How it relates to SIMY: Work and Codex are doers that finish the one task you assign. SIMY is the manager that picks up work from meetings and chats, runs the doer, and returns only decisions. SIMY uses Codex as its doer, so you use them together, not one instead of the other.

• Is it used for training? Business and Enterprise data is not used for model training by default. On personal plans it is on by default, so go to Settings → Data Controls and turn off "Improve the model for everyone."
• Main leak paths: shared chat links, a training setting left on, permissions given to plugins, and the file scope allowed for local runs. Settle these four first.
• Work permissions: Work Cloud, Work Local and Codex Local can be configured separately. On business plans, follow your workspace admin's policy.
• Five items for your guidelines: which information may be entered, training settings, share link rules, allowed scope for plugins and local runs, and who is responsible for checking output.
• SIMY's approach: Gmail stops at saving drafts. Posting to Slack and writing to Notion require your approval. When suggestions are created from conversation history on your computer, conversation content is not stored on our servers. Details are on our Security and Privacy Policy pages.