ChatGPT Dots for Nonprofits: Which Seat Gets One and What to Lock Down First

Scottship Solutions helps nonprofits decide where AI agents belong in their workflows, the core of our AI and automation for nonprofits work, and this guide covers ChatGPT dots for nonprofits as of September 30, 2026: the always-on agents OpenAI launched on September 29 that run on their own cloud computer and keep working between conversations. The seat that carries a dot is not the discounted one: OpenAI’s help center confirms the Business Standard seat at $8 per user per month billed annually does not include a dot, Business Premium seats are excluded from the nonprofit discount, Pro has no discount, and Enterprise access is an off-by-default beta. In the 5 Levels of AI framework, that makes a dot Level 4 behavior on a Level 1 license, which is why the governance decisions below matter more than the setup steps.

Scottship’s AI and automation work is led by Parker Davis, IBM AI Product Manager and Claude Certified Architect. Our earlier process-automation work for Carousel Child Advocacy Center returned more than 750 staff hours a year, and that is the lens for this guide: the recurring job comes first, the tool second. Nothing in this post describes a dots outcome for any client, because the feature is one day old.

Every OpenAI fact below is stamped as of September 30, 2026 and tied to the OpenAI article it came from. That matters here more than usual. The three dots help articles all carried update stamps from the previous twelve hours when we read them, so check the dated Sources before you budget.

What You’ll Learn

1. What are ChatGPT dots?
2. Which ChatGPT plans include dots?
3. Where do ChatGPT dots sit in the 5 Levels of AI framework?
4. What can a ChatGPT dot do while nobody is watching it?
5. What data does a ChatGPT dot remember, and can you delete it?
6. Is it safe to connect donor or client data to a ChatGPT dot?
7. What should a nonprofit decide before turning on a dot?
8. Should a nonprofit turn on ChatGPT dots?
9. Frequently Asked Questions
10. Your Next Steps

What are ChatGPT dots?

A dot is an always-on agent in ChatGPT that can take on ongoing responsibility and keep making progress between conversations, in OpenAI’s own words from the help article “Getting started with your dot.” It is powered by GPT-6 Astra, has its own cloud computer, works across the apps you choose to connect, and remembers context for ongoing work. OpenAI launched dots on September 29, 2026 at DevDay. Headlines call them ChatGPT Dots, OpenAI Dots, or always-on AI agents; OpenAI’s documentation uses lowercase dots, and so does this post.

For a 10-to-50-person nonprofit, the plain-language version is this: a dot is a standing assistant that does one recurring job between meetings, not a chatbot you prompt. It is the always-on AI assistant that nonprofit software vendors describe in their own product pitches, except it lives inside ChatGPT rather than inside your donor database. That location is the whole story of this post.

You create a dot in the ChatGPT desktop app or on desktop web. Dots are not supported on mobile web as of September 30, 2026, and OpenAI has not given a date for creating them on mobile. Once created, a dot works in ChatGPT and, where an admin allows it, in Microsoft Teams and other team-chat apps. Your laptop does not need to stay on, because the dot runs on its own cloud computer and access to your local machine is optional and starts turned off.

Which ChatGPT plans include dots?

The seat the nonprofit discount buys is the one that does not carry a dot. OpenAI’s nonprofit help article discounts Business Standard seats for verified nonprofits and states in plain terms that Business Premium seats are not eligible for the discount. The dots help article puts dots on Pro and Business Premium, with Enterprise as an admin-enabled beta. Free, Go, and Plus are not on the list.

On the plans that carry one, your first dot is included at no extra cost, with extended limits for the first month, per OpenAI’s “Getting started with your dot” article. The plan also includes an allowance for deeper work. OpenAI says you will be able to add more dots in the future and has published no price or allowance for them yet.

Discount mechanics, Goodstack verification, and the Business versus Enterprise comparison live in our guide to ChatGPT for nonprofits, so this post does not restate them.

One correction, because AI engines are already getting this wrong. At least one chatbot answer we reviewed implied that a nonprofit discount could apply to the dot-carrying seat. It cannot. OpenAI’s wording is that nonprofit discounts apply only to Business Standard seats, and if your workspace has both Standard and Premium seats, only the Standard seats receive the discount.

Where do ChatGPT dots sit in the 5 Levels of AI framework?

In the 5 Levels of AI framework, Level 1 is AI Tools: off-the-shelf products used as-is, and ChatGPT is the first example we name. Level 4 is AI Agents: autonomous multi-step execution, AI that acts rather than assists. Most nonprofits are stuck on Levels 1 and 2. A dot is Level 4 behavior shipped inside a Level 1 product, and that collision is the reason governance matters more than setup.

The framework predicts this exact problem. A Level 1 tool arrives through a per-seat license that any staff member with a card can buy, while Level 4 behavior normally arrives through a scoped build with an owner, a data boundary, and a review cadence. Dots deliver the second through the purchasing path of the first. The capability shows up before the policy does.

Here is the decision rule we use. A Business Premium seat costs about $1,200 per user per year at the annual rate, so a dot beats a Level 2 Zapier or Make workflow only when the recurring job changes week to week and needs judgment: reading a grant portal that reorganizes its pages, or summarizing whatever arrived in a program inbox. When the job is a fixed trigger and action, a new donation creating a thank-you task for example, Level 2 automation does it for a fraction of the cost and with no memory to govern.

A nonprofit that wants Level 4 behavior with a governed build instead of a personal seat is describing our AI agents for nonprofits work, where the agent runs on the organization’s own data. That build is the natural next step once a dot pilot proves the recurring job is real. Give it a written scope from day one.

What can a ChatGPT dot do while nobody is watching it?

Dots are ChatGPT’s own agent feature, which makes them the literal answer to the question nonprofit leaders have been typing as “ChatGPT agents for nonprofits.” What a dot does unattended depends on which of two modes it is in: proactive research or a delegated task.

Proactive research is read-only. OpenAI’s privacy and safety FAQ states that in this mode the dot can read information from permitted connected sources and save private notes, but its research tools cannot send messages to other people, change content through plugins, or control a browser or computer. The dot also forms memories from what it reads, even when you have not asked a question about it.

Delegated tasks are where actions happen, and Custom Rules decide how. For each supported action you choose one of four behaviors OpenAI names in the setup article: take action without asking, take action if pre-approved, ask before taking action, or hand off to you. On top of that sit OpenAI’s other safeguard layers: model safeguards, plugin permissions, Auto-review, and safety monitoring.

The most sensitive actions, changing a password or transferring money, require you to take over. OpenAI also states that your approval cannot override core safety requirements. Read those two lines together and the design intent is clear: the dot drafts and researches on its own, and a person stays in the loop for anything that moves money or credentials.

Three recurring jobs shaped for a nonprofit, offered as design patterns rather than client results: watching a grant portal for deadline changes and drafting a summary; assembling a board packet outline from a shared drive; triaging a program inbox into draft replies that a human sends. Each one is read-heavy, produces a draft, and leaves the send to a person. That is the shape to start with.

What data does a ChatGPT dot remember, and can you delete it?

Three facts from OpenAI’s privacy and safety FAQ, as of September 30, 2026, that no nonprofit publisher has stated. First, you cannot view, delete, or directly modify individual dot memories, including specific details that enter the dot’s context from plugins. Second, disconnecting an app stops new access but does not delete information the dot has already built into its context. Third, the only full reset is deleting the dot.

Deleting is not a clean slate either. Reset removes the dot with its conversations, saved memories, and scheduled tasks, but files, Codex threads, and ChatGPT conversations the dot created are stored separately and survive it. Turning off Memory in ChatGPT stops new sharing with the dot and does not delete what it has already received.

On training, OpenAI states that for ChatGPT Business, Enterprise, and Edu workspaces it does not use your data to train its models by default, and that human review may occur in limited circumstances, including safety-related cases, even when model improvement is turned off. Pro is not named in that workspace statement. A staff member on a personal Pro plan manages their own settings, outside your admin’s view.

The Business Associate Agreement (BAA) point carries over from our ChatGPT guide unchanged: OpenAI does not offer BAAs for ChatGPT Business and directs anyone who needs one to Enterprise. A health-adjacent nonprofit therefore cannot run a dot against protected health information anywhere except the Enterprise beta, and only after the admin turns it on.

Is it safe to connect donor or client data to a ChatGPT dot?

Not on a first dot. Four categories stay off it: the donor CRM, client case notes, HR files, and finance systems. The memory facts above are the reason. Once a dot has read a donor record or a case note, there is no per-memory delete, and disconnecting the app does not take it back.

Workspace dynamics matter too. Only a dot’s owner can direct it, and other people’s direct messages or mentions do not start work for it, but when a dot posts in a channel every channel member can see its messages. A dot that summarizes a program inbox into a shared Teams channel has just shared whatever it read with everyone in that channel.

The cloud computer changes the perimeter. A dot runs on its own machine in OpenAI’s cloud, not on the staff member’s laptop, and access to the local computer is optional and off by default. The device policies your IT team enforces on laptops do not reach that cloud machine on their own, so the connections you approve are the boundary. Two Enterprise details from the workspace admin article reinforce that: enabling dots does not grant access to every app or website, and Enterprise model controls and default model settings do not apply to dots.

Two beta limits appear in OpenAI’s learning center rather than in the three help-center articles above. OpenAI’s learning-center article on local computer access for dots states that during the Enterprise beta, dots do not support data residency or inference residency, and that dots are unavailable for FedRAMP workspaces, workspaces with EKM, and workspaces with inference residency set to AE (UAE). The same article adds that HIPAA workspaces can participate if they meet the other eligibility requirements. If your nonprofit holds government contracts or sits under residency requirements, put that question to OpenAI before a pilot, and treat the answer as unsettled until you have it in writing.

OpenAI’s own caveat on prompt injection deserves a direct quote: its protections “help reduce the risk of malicious instructions causing an unwanted action, but they do not eliminate it.” OpenAI shipped dots with read-only proactive research and layered safeguards on top of Custom Rules, and that is the posture we would echo in a pilot. Start read-only, expand on evidence, and keep the four categories above off the dot until the first 30 days have shown you what it actually reads.

What should a nonprofit decide before turning on a dot?

Six decisions, in plain language, before the first dot goes live. Write each one down. A dot with no written owner and no written boundary is the unmanaged-connector problem wearing a new name, and the same pattern that catches organizations with connectors will catch them here.

1. Who is allowed to create a dot. Name the roles, not the people, and default everyone else to no.
2. Which apps it may connect to. An allow list, written before the first connection, with the four sensitive categories excluded.
3. Read-only versus write access, per system. Decide this for each connected app, not for the dot as a whole.
4. The Custom Rules starter set. The five rules below, adopted as written for month one.
5. Who owns the dot and its memory when the staff member leaves. Since individual memories cannot be viewed or deleted, the answer is a named successor or a Reset on the exit date.
6. A review date after the first 30 days. Read what the dot touched, then decide what to expand.

The Custom Rules starter set for a nonprofit’s first month is short enough to paste into a board memo. It errs toward drafting and away from sending, and it keeps the systems that hold personal data out of reach entirely.

• Never send email in month one; every message is a draft a person sends.
• Never purchase anything.
• Require approval before sharing anything outside the organization’s domain.
• Read-only on the donor CRM and finance systems, if they are connected at all.
• No connection to HR or client-services drives.

On Enterprise, the admin path is short. Dots access is off by default, the permission is named “Use dots (Beta)”, and it is granted by role, so scope it to a pilot custom role rather than the workspace default. Local computer access and custom rules for dots are separate permissions that also start off, and the cloud controls for browser use, network access, and computer use sit alongside them. One trap from the same article: a role that grants access can still allow a member to use dots even if another role has the permission turned off, so check every role a pilot member holds.

On Business, the question is open. OpenAI’s workspace management article covers Enterprise in detail and never mentions Business workspaces, and as of September 30, 2026 we have not found documentation either way on whether a Business workspace owner can restrict dots for Premium seat holders. A nonprofit on Business should ask OpenAI that question in writing before buying Premium seats for a pilot.

Two earlier posts cover the policy side in depth: our read on AI agent governance after a production agent failure, and our guide to nonprofit AI governance built around auditing the tools staff already use. Both apply to dots without modification.

Should a nonprofit turn on ChatGPT dots?

The verdict follows the seat you hold. A nonprofit on discounted Business Standard seats should not buy Premium seats for a dot until it has a defined recurring job and the rules above written down. A nonprofit already on Enterprise can run a scoped, read-only pilot for one custom role and review it at 30 days. A staff member on a personal Pro plan should not connect organizational data to a dot at all.

What OpenAI has not published as of September 30, 2026: the price of additional dots, the deeper-work allowance after the first month, timing for creating dots on mobile, and any change to Business Premium nonprofit pricing. Those four gaps are the reason to pilot small and budget nothing beyond the seats you can already justify. This section is dated so it ages honestly.

At Scottship Solutions, we help nonprofits put AI agents where they earn their keep and keep them out of the systems where they do not belong. If your team is weighing a dot, an agent build, or a governance policy for either, we can walk you through it.

Frequently Asked Questions

How much does a ChatGPT dot cost a nonprofit?

As of September 30, 2026, the first dot is included at no extra cost on the plans that carry one, with extended limits for the first month. The seat is the cost: Business Premium lists at $100 per user per month billed annually or $125 monthly and is excluded from the nonprofit discount, while the discounted Business Standard seat at $8 per user per month billed annually does not include a dot. OpenAI has not published a price for additional dots.

Does the OpenAI nonprofit discount apply to ChatGPT dots?

No. OpenAI’s nonprofit help article states that nonprofit discounts apply only to Business Standard seats and that Premium seats are not eligible, and Standard seats do not include a dot. The nonprofit program covers Business Standard seats and negotiated Enterprise pricing, so Pro carries no nonprofit discount. Discount mechanics and Goodstack verification are covered in our ChatGPT for nonprofits guide.

What is the difference between a ChatGPT dot and ChatGPT agent mode?

Agent mode runs inside a conversation you are having and works while that session is open. A dot keeps making progress between conversations on its own cloud computer, remembers context across the ongoing work you give it, and is limited to read-only tools while it researches proactively. Agent mode is a session; a dot is a standing assignment.

Can a ChatGPT dot send email or post messages on its own?

Not during proactive research, which is read-only: OpenAI states the research tools cannot send messages to other people. Sending only happens inside a delegated task, where Custom Rules set whether the dot acts without asking, acts if pre-approved, asks first, or hands off to you. At Scottship, we recommend a first-month rule of no outbound email at all, with every message handed to a person who sends it.

Are ChatGPT dots available on ChatGPT Enterprise for nonprofits?

Yes, as a beta that is off by default. A workspace owner enables the “Use dots (Beta)” permission and scopes it by role, and local computer access and custom rules stay off until they are separately enabled. Enterprise is also where OpenAI directs BAA requests and the only tier where a nonprofit discount, negotiated through sales at up to 75 percent, and a dot can coexist. Scottship’s recommendation for Enterprise nonprofits is a single pilot role, read-only, reviewed at 30 days.

Your Next Steps

1. Check your seat type: Open your ChatGPT workspace billing page and confirm whether you hold Standard, Premium, or Enterprise seats. Standard means no dot.
2. Name the recurring job: Write one sentence describing the job a dot would own, and test it against the Level 2 rule: if a fixed trigger and action covers it, use Zapier or Make instead.
3. Write the six decisions: Owner, allowed apps, read or write per system, Custom Rules, succession, and the 30-day review date, on one page, before any seat is purchased.
4. Ask OpenAI the Business question: If you are on Business, get a written answer on whether a workspace owner can restrict dots for Premium seat holders.
5. Scope the pilot to one role: On Enterprise, enable “Use dots (Beta)” for a single custom role, leave local computer access off, and check every role each pilot member holds.

Sources

• OpenAI Help Center, Getting started with your dot: definition, GPT-6 Astra, cloud computer, first dot included at no extra cost, rollout by plan and region, desktop-only creation, Custom Rules behaviors, Reset scope
• OpenAI Help Center, Dots privacy, security, and safety FAQs: read-only proactive research, memory cannot be viewed or deleted individually, disconnect and delete behavior, training defaults, human review, safeguard layers, prompt injection caveat
• OpenAI Help Center, Manage dots in ChatGPT workspaces: Enterprise-only scope, off by default, Use dots (Beta) permission, role behavior, cloud controls, channel visibility, model controls do not apply
• OpenAI Help Center, OpenAI for Nonprofits: Business Standard discount, Premium seats not eligible, Enterprise discount up to 75 percent, no BAA on Business
• OpenAI, ChatGPT Business pricing: Premium seat at $100 per month billed annually or $125 monthly
• OpenAI, Introducing dots: launch announcement, September 29, 2026
• OpenAI Learning Center, Local computer access for Work Cloud and dots: Enterprise beta data residency and inference residency limits, FedRAMP, EKM, and UAE inference residency exclusions, HIPAA workspace eligibility

Written by

Parker Davis

System Architect at Scottship Solutions

Parker works hands-on with nonprofit and small business leaders on IT strategy, cybersecurity, AI implementation, and cloud architecture. He leads the work at Scottship Solutions that helps mission-driven organizations build reliable, secure technology infrastructure.

Certifications

IBM AI Product Manager • Lean Six Sigma Yellow Belt • TOGAF Standard • Claude Certified Architect

Industries Served

Human Services, Healthcare & Community Health, Arts & Culture, Education & Youth Development, Faith-Based, Foundations & Grantmakers, Child Advocacy