ChatGPT: Advanced

Advanced

At this level ChatGPT is part of daily work. The gains now come from feeding it your real data through connected apps, packaging repeatable work as a custom GPT, and knowing exactly where your data goes. The discipline underneath all of it is verification: the app gives you traceable evidence, not automated tests.

Apps and connectors: bring your own data

Apps (the feature was renamed from connectors on 17 December 2025) link ChatGPT to the tools you already use, so it can read (and increasingly write) from inside a chat. Coverage includes Google Drive, SharePoint/OneDrive, Dropbox, Box, Gmail and Outlook, Google and Microsoft Calendar, GitHub, Slack, Teams, Linear, HubSpot, and Asana, among others.

• Read: “Find last quarter’s pricing doc in Drive and summarize the changes.”
• Write: newer actions can draft an email, create a doc or sheet, or schedule a meeting (rolling out through 2026).

Note

On Business, Enterprise, and Edu, Company knowledge answers from your connected apps with citations back to the source document. Note: write actions are disabled in Company-knowledge mode, so select the app directly when you need it to write.

Regional availability varies; see the tenant checklist at the end of this page before you build a workflow on a specific app.

Build a GPT: package the repeatable work

A GPT is a no-code custom assistant: a set of instructions, optional uploaded knowledge files, and capabilities (web, images, data analysis). Build one when you find yourself pasting the same setup into chat after chat.

1. Open the GPT builder and write its instructions: its standing role and rules.
2. Upload reference files it should always know.
3. Optionally add Actions (REST/OpenAPI calls) so it can reach an external service.
4. Share it by link or to your workspace.

Tip

Building a GPT needs a paid plan, but anyone, including Free users, can use one you share. That makes a shared GPT a clean way to hand a teammate a ready-made workflow.

Preview code and drafts inline

On GPT-5.5 and later models, long drafts open in inline writing blocks and code blocks (these replaced Canvas, which OpenAI removed from GPT-5.5 on 28 May 2026). Code blocks can preview and run output directly (HTML, React, SVG, Mermaid, Vega, and runnable Python with console output), so you can see the result instead of guessing.

Write a Mermaid flowchart of this onboarding process, then show me the rendered diagram.

Tip

The Sources panel (the book icon under a response) shows whether an answer drew on memory, past chats, files, or an App, and why. Open it whenever an answer needs to be trustworthy, and prefer connectors or Company knowledge so claims carry links you can check.

Control your data

This matters most for a Swiss SME handling client data.

Control	What to know
Training	OpenAI does not train on Business, Enterprise, or Edu data. On Free/Plus/Pro, data may be used to improve models unless you toggle that off.
Data residency	Europe (EEA + Switzerland) storage is supported on eligible business plans; in-region inference for Europe was added in January 2026 and is arranged through sales, not self-serve.
Lockdown Mode	An opt-in mode that blocks the exfiltration paths prompt-injection attacks use. It disables live web browsing (cached content only), web images, deep research, and agent mode while keeping the model usable.
Admin controls	SSO/SCIM, retention settings, connector allow-listing, and remote sign-out on Business/Enterprise.

Verify on your tenant and region before you rely on it

Almost everything on this page varies by plan, tenant, and region. Several apps and connectors roll out slower in the EEA, Switzerland, and the UK; data residency and in-region inference are limited to eligible business plans; and personal Plus/Pro is not a governance boundary. Before building a client workflow: put client or tenant data on Business or Enterprise with CH/EEA data residency configured, then confirm the specific app, feature, and control on the actual target tenant, not in a demo account.

You’re done when…

• ✅ You’ve connected an App and pulled real data into a chat
• ✅ You’ve built and shared a GPT
• ✅ You’ve previewed runnable output in a code block
• ✅ You can explain where your plan’s data is stored and trained

Next step

Ready to hand ChatGPT a whole task and let it work, under supervision?

ChatGPT: Expert The agent, scheduled tasks, and where the consumer app stops.