ThriveAI vs OpenAI custom GPTs and ChatGPT Business
OpenAI plans to retire custom GPTs on December 11, 2026 and move them to plugins, so a team that relies on a GPT has a choice to make before then. This page gives you the facts that decide it: what the retirement changes, how ChatGPT Business charges in Canada, where each option keeps your data, and when each one is the better buy.
The short answer
- Pick ChatGPT Business when your team needs help writing, researching and answering questions from documents, and OpenAI's ready-made apps cover the systems you use. It charges per person per month, as the ChatGPT pricing page shows, and you sign up yourself.
- Pick ThriveAI when the work runs on your ERP, drawings and email, and you want those records kept on your own server in Canada. ThriveAI works with your team until the tool is in daily use.
- Use both when your staff want ChatGPT for everyday writing and research, and one job in the business needs a system built on your own records.
What changes when custom GPTs retire
OpenAI plans to retire custom GPTs on every ChatGPT plan and move them to plugins. The change appears in the ChatGPT release notes under September 11, 2026. OpenAI's custom GPT retirement FAQ gives December 11, 2026 as the scheduled retirement date, when custom GPTs stop running and leave the GPT directory. Existing GPTs keep working until then.
A plugin packages skills, connected apps and app templates, as OpenAI's plugins article explains. When a GPT migrates, its instructions become a skill, its knowledge files become reference files, and its connected apps come along as apps. The retirement FAQ lists what stays behind:
- Custom actions. An action does not transfer. OpenAI says rebuilding that connection “may require a custom MCP server and technical setup.”
- The selected model. The plugin does not keep the model that you picked for the GPT.
- Sharing settings. The plugin starts private, and the people who used the GPT do not get access automatically.
- Conversations. Past GPT conversations stay with the original GPT until it retires.
A custom action let a GPT call another company's software. Setting one up took that service's API details, its sign-in method and an OpenAPI schema, according to OpenAI's article on GPT actions. An MCP server does the same job for plugins: it is a small service, run by you or a vendor, that lets ChatGPT read from one of your systems or write to it.
The planned dates
For Enterprise workspaces, the retirement FAQ lists these planned milestones and says that the dates may change.
| Date | Planned milestone |
|---|---|
| September 11, 2026 | Notice to workspace admins. |
| September 22, 2026 | Target date for the migration tool. OpenAI says a target date does not guarantee that every workspace gets it that day. |
| October 26, 2026 | Planned end of creating new custom GPTs. |
| December 11, 2026 | Scheduled retirement. Custom GPTs stop running. |
Personal accounts have already lost the ability to build GPTs. OpenAI's article on creating GPTs says Free, Go, Plus and Pro accounts can no longer create or publish new GPTs. People in Business, Enterprise and Edu workspaces can create them if their workspace settings allow it.
If a GPT runs part of your work today, list the GPTs your team relies on and note which ones use custom actions. A GPT that holds only instructions and files migrates through OpenAI's tool. A GPT that reaches your ERP or another system through an action needs that connection rebuilt and tested before December 11.
What ChatGPT Business gives you
ChatGPT Business is OpenAI's self-serve plan for teams of 2 to 200 people. It sells two kinds of seat, each charged per user per month. On September 23, 2026, the ChatGPT pricing page showed the prices in Canadian dollars to a visitor in Canada:
- Standard seat. Per user per month, billed annually or monthly.
- Premium seat. Per user per month at a higher rate, billed annually or monthly. Premium gives five times the usage of Standard, with no 5-hour limit.
OpenAI's billing and seats article lists the seat prices in US dollars. Every workspace needs at least two seats, and you can mix the two seat types. The bill grows with the number of seats, and annual billing costs less per seat than monthly billing. For how seats compare with other ways to pay for AI, see the cost of AI.
The pricing page lists these features for Business:
- Apps that connect to Google Workspace, Slack, GitHub, Microsoft 365 and other services.
- Workspace agents, company knowledge, and developer mode (in beta) for custom apps.
- SAML single sign-on.
- No training on Business data by default.
Contract terms
Self-serve Business is paid by card. The billing article says invoices, purchase orders and net terms are not available on it, and it sends companies that need invoicing to Enterprise. On an annual plan, seats you add stay committed until the end of the term. Seat reductions take effect at the next billing cycle, and charges are generally non-refundable.
What Enterprise adds
ChatGPT Enterprise has custom pricing through OpenAI's sales team. OpenAI publishes no Enterprise price and no minimum seat count. The pricing page marks these features as Enterprise only: data residency in 10 regions including Canada, SCIM, Enterprise Key Management, role-based access controls, the Compliance API, IP allowlisting, a connector registry, dedicated onboarding and a custom security review. The ISO 27001, 27017, 27018 and 27701 row is also marked for Enterprise only. Enterprise lists 24/7 priority support, SLAs, custom legal terms, invoicing and volume discounts.
Workspace agents
Workspace agents let ChatGPT Business run work without someone typing. OpenAI's workspace agents article says you can run an agent on a schedule, trigger it through an API or use it in Slack. An agent can connect apps, custom MCP servers, skills and files of up to 512 MB each, with 10 GB per agent. Agents are off by default on Enterprise. An API trigger returns an acknowledgement with no run ID. OpenAI says the agent's response cannot currently be retrieved through the API, so the system that started the agent gets no answer back from that call.
Side-by-side comparison
The ChatGPT column describes Business unless a cell says otherwise. Every OpenAI fact links to OpenAI's own page, checked on September 23, 2026.
| Topic | ThriveAI | ChatGPT Business |
|---|---|---|
| What you get | A private AI system built on your company's records, and hands-on work with your team until it is in daily use. | A shared ChatGPT workspace with apps, plugins and workspace agents that your own admins set up. |
| How it charges | Each project is quoted on its own, and no rates are published. Stage one is a working prototype for one job, at a fixed price. | Per user per month for Standard and Premium seats, with at least two seats (pricing page). Self-serve plans are paid by card (billing article). |
| Where your data is stored | The platform is designed to keep your data on its own server in Canada that only your company uses. | Business has no data residency option. Storage at rest in Canada is open to new Enterprise and Edu customers (residency article). |
| Where the model runs | You choose a model on your own server, or a hosted model under a written agreement that the provider keeps nothing after answering. A hosted model may process requests outside Canada. | Inference residency is offered in Europe, the United States and the UAE. Canada is not on the list (residency article). |
| Your systems | Reads your ERP, email, drawings, shared drive and spreadsheets, including older ERP versions on your own server. Every connection only reads data. | Apps for Google Workspace, Slack, GitHub, Microsoft 365 and others (pricing page). Any other system needs an MCP server that you supply and check for safety (developer mode article). |
| Documents it reads | Brings what your ERP, email, drawings and spreadsheets hold into one database that belongs to your company. Each answer links to the record it came from. | A GPT holds up to 20 knowledge files of up to 512 MB each (GPT article). A workspace agent holds up to 10 GB of files (agents article). |
| Work that runs by itself | A task you repeat every week runs by itself. Anything going to a customer or a supplier waits as a draft until someone on your team approves it. | Workspace agents run on a schedule, from an API trigger or in Slack (agents article). |
| Who sets it up | ThriveAI works on site with your team, and Derik Lawlis leads the project. | Your admins configure app connections, publish apps and assign access (plugins article). Dedicated onboarding is an Enterprise feature (pricing page). |
| Security and compliance | No audited certification. Each security measure is written down so your IT lead or outside security adviser can review it. | SAML single sign-on is included. SCIM, role-based access controls, the Compliance API and the ISO 27001 family are marked Enterprise only (pricing page). |
| What you keep | You keep your data, and everything the system recorded exports in full in a documented format. A stage-one prototype stays with you. | Custom GPTs retire on December 11, 2026. Migration to a plugin does not carry over custom actions, the selected model, sharing settings or past conversations (retirement FAQ). |
| How you start | Stage one delivers a working prototype for one job in weeks. | Self-serve sign-up, paid by card (billing article). |
Where your data sits
For a manufacturer, the files at stake are customer drawings, supplier prices and ERP exports. OpenAI treats two questions separately: where content is stored at rest (data residency) and where the model processes it (inference residency). Its data residency article sets out both:
- Storage in Canada. New ChatGPT Enterprise and Edu customers, and eligible API customers, can choose data residency, and Canada is one of 10 supported regions. It costs nothing extra on Enterprise and Edu. Business is not eligible.
- Processing in Canada. Inference residency covers Europe, the United States and the UAE. With Canadian data residency, the model can still process your content outside Canada.
- What can leave the region. Data sent to apps, MCP servers or web search, some processing steps, and workspace metadata may be handled outside the chosen region. Third-party GPTs are not eligible for residency.
On retention, OpenAI's enterprise privacy page says deleted or unsaved Business conversations are removed within 30 days. Business admins can view, export and delete members' conversations, and OpenAI signs a data processing agreement for Business, Enterprise and the API. Files uploaded as GPT knowledge stay until the GPT is deleted, and OpenAI then deletes them within 30 days, according to its chat and file retention article.
If a developer builds on the OpenAI API instead, the API data guide says the Canada region supports data storage and does not support processing. Using it requires approval for modified abuse monitoring or zero data retention. Data residency endpoints cost more on eligible models released on or after March 5, 2026. By default, abuse-monitoring logs are kept for up to 30 days.
The platform ThriveAI builds on is designed to keep each client's data on its own server in Canada. You choose the model that reads it: one that runs on that server, or a hosted model under a written agreement that the provider keeps nothing after answering. A hosted model may process requests outside Canada, so the contract names the model and its service tier.
What you build yourself with ChatGPT
OpenAI's ready-made apps cover services such as Google Workspace, Slack, GitHub and Microsoft 365. If your ERP or another system has no ready-made app, someone on your side supplies the connection. OpenAI's developer mode article describes what that takes:
- You supply an MCP server endpoint and its authentication, and OpenAI says you are responsible for verifying that the server is safe.
- With OAuth sign-in, the server needs to issue refresh tokens, or people may have to sign in again.
- Only admins and owners can publish an app. On Business, OpenAI says a published app cannot be updated at launch, so a change means recreating and republishing it.
- Full write support, which lets ChatGPT change data in the connected system, is in beta.
App templates inside plugins work the same way. The plugins article says an administrator may need to configure the provider connection, publish a workspace app and assign access before anyone can use one.
Someone also has to maintain that server after it is built. If you hire for it, Job Bank lists a median wage of $56.73 an hour for AI software engineers (NOC 21231) in the Ottawa region. The median in the Montréal region is $52.88 an hour. Both figures come from the 2023-2024 Labour Force Survey. At 2,080 hours a year, the two medians come to about $118,000 and $110,000 in wages.
Help from OpenAI's own engineers
OpenAI also places engineers inside companies. It launched the OpenAI Deployment Company on May 11, 2026 to put forward deployed engineers inside organizations. OpenAI says a typical engagement starts with a diagnostic and then covers a small number of priority workflows, and it publishes no minimum engagement size or price. Its Enterprise page says eligible customers get hands-on help with implementation, training and change management. AI advisors work with select Enterprise customers that roll ChatGPT out across the whole organization.
When to pick each
Pick ChatGPT Business when
- Your team wants help writing, summarizing, researching and answering questions from documents.
- Your work lives in Google Workspace, Microsoft 365 or Slack, which have ready-made apps.
- A person reads every answer before it goes anywhere, so a wrong answer costs little.
- You have between 2 and 200 people, and paying per seat by card suits you.
- You have no requirement to store data in Canada. If you do, Canadian storage is available only on ChatGPT Enterprise, for new customers.
- You want to start this week without outside help.
If this list describes your needs, ChatGPT Business is the better buy.
Pick ThriveAI when
- The job runs on your ERP records, drawings, purchase history or email, and a wrong number costs money.
- Your ERP or other key systems have no ready-made ChatGPT app.
- You want your records kept on your own server in Canada, with the model named in the contract.
- You want every answer to show the record it came from, and every draft to wait for a person's approval.
- Nobody on your team has time to build and maintain an MCP server.
- A custom GPT with a custom action runs part of your work, and it needs rebuilding before December 11, 2026.
- You want someone working with your team on site, in French or English.
Using both
The two options do different jobs, and one company can run both. Your staff can keep ChatGPT Business for writing and research. A ThriveAI system then handles one job that runs on your own records, such as order entry from emailed PDFs or supplier follow-ups. Write down which files go where: customer drawings, costing sheets and ERP exports go into the private system, and general questions go to ChatGPT. For the jobs a private system takes on at each desk in a plant, see AI for manufacturing, desk by desk.
How a ThriveAI project works
ThriveAI is an AI engineering company in Ottawa that builds private AI systems for manufacturers and distributors in Ontario and Quebec, on their own data, and works hands on with your team. Derik Lawlis, the founder, leads every project and stays close to the build. For the company's background, see About ThriveAI.
- Stage one. ThriveAI builds a working prototype for the one job that costs you the most, at a fixed price, in weeks. If you stop there, you keep it.
- Stage two. The full build connects the systems you already run. You decide whether to start it after you have seen stage one working.
- Stage three. ThriveAI keeps adding to the tools as your business changes and as AI models get better.
Every project is quoted on its own, and no rates are published.
Working with Derik to build these AI agents has been fun, very worthwhile, with a huge ROI for my team.
Fahd Alhattab, Founder, Unicorn Labs.