On July 21, OpenAI introduced the ChatGPT for small business program, an initiative that brings together training, events, practical guides, and resources from participating business-software companies. The announcement also promotes ChatGPT Work, a product capability for multi-step tasks.
Each part answers a different need. The program helps a business learn where ChatGPT may be useful, ChatGPT Work carries out tasks across connected information, and ChatGPT Business provides controls for a managed team workspace. A business with no defined workflow should start with training; one with a repeated, reviewable task can test Work; and a team rollout should include a Business workspace review before employees connect company data.
The program provides training, events, guides, and partner resources
OpenAI describes the program as a small-business adoption initiative with four parts. It includes hands-on virtual training; in-person small-business AI academies in the United States; guides, customer stories, uploadable interactive guides, and short videos; and partner resources such as plugins, skills, offers, and webinars. OpenAI names Dropbox, Shopify, Intuit, Slack, Atlassian, and Wix among the participating companies.
OpenAI asks businesses to sign up for updates and future opportunities, while the small-business landing page links to current sessions and resources. The announcement does not state a fee for joining the program or say that every event, partner offer, connected product, or service is free. OpenAI also says questions, local events, and surveys will inform future product development and resources.
ChatGPT Work handles multi-step tasks across connected files and applications
OpenAI describes ChatGPT Work as an agent for multi-step tasks. In this context, an agent is software that can continue through several parts of a task instead of responding to one prompt at a time. OpenAI says ChatGPT Work can connect with business files and applications and use memory about how a person works.
That capability is relevant when a business has a repeated task with clear inputs and a reviewable output. A team might use it to gather information from approved sources, prepare a draft, compare records, or assemble material for a person to review. The task still needs defined sources and a clear point where an employee takes over.
The announcement says ChatGPT Work and GPT-5.6 are available to small businesses today. It separately says GPT-5.6 is available to businesses of any size and across all subscription plans. The announcement does not say that every plan receives identical connectors, limits, credits, admin controls, or data terms, so a business should verify the current details for the subscription it intends to use.
ChatGPT Business adds controls for a managed team workspace
OpenAI presents ChatGPT Business as the option for companies that need a shared workspace with team administration. On the program landing page, OpenAI lists an admin console, user roles, SAML single sign-on, multi-factor authentication, encryption, and no training on business data by default as ChatGPT Business controls.
Those controls belong to the Business workspace claim and should not be assumed for every plan simply because GPT-5.6 or a product capability is broadly available. Work concerns what the agent can do across a task; Business concerns how an organization manages a team workspace. A company evaluating access should check the current plan documentation and its own requirements for identity, user removal, data handling, and application connections.
The right first layer depends on how clearly the business can define the work
A business still deciding where AI is useful should begin with the program’s training, examples, or events. The goal at this stage is to identify one repeated task, the information it uses, the person responsible for the result, and the mistake that would make the test unacceptable. Buying broader access or connecting applications does not resolve an unclear workflow.
A business that already knows the task can evaluate ChatGPT Work with a narrow source boundary. BaristaLabs recommends choosing work that occurs often, produces an output a person can inspect, and can be corrected before it changes a customer promise, public page, payment, or system of record. Our guide to choosing a first AI workflow by what you can safely undo explains how to make that choice before expanding permissions.
A team preparing to give several employees access should review ChatGPT Business and its current workspace terms before rollout. That review should cover who can add and remove users, which data employees may connect, what administrators can see, and how the company will handle departures or role changes. Product availability alone does not answer those operating questions.
Use one current session to define one workflow before connecting more data
As of July 21, OpenAI’s landing page lists sessions on sales use for July 30, ChatGPT Work for small businesses on August 6, Shopify on August 19, QuickBooks on August 25, and Codex for small-business developers on August 27. Event schedules can change, so check the current page before planning around a date.
Choose the session closest to a task your business already repeats. Before attending, write down the task’s trigger, the minimum files or applications needed, the output an employee will review, and the actions the system cannot take during the first test. After the session, decide whether the task is specific enough to test; if it is not, keep learning rather than adding connections.
For the first test, connect only the sources needed for that task and use read-only access where the product and workflow allow it. Keep sending, publishing, paying, approving, and updating systems under human control until the team has reviewed representative results and recorded the corrections. The BaristaLabs guide to AI workflow controls covers the practical checks around sources, permissions, review, and action boundaries.
Judge the test by what employees can verify: whether the agent used the right sources, how often reviewers corrected the output, how much review time the task required, and which cases stayed manual. That evidence supports a later decision to expand, revise, or stop.
Sources
- OpenAI: Introducing the ChatGPT for small business program
- OpenAI: ChatGPT for small business
- OpenAI: ChatGPT for your most ambitious work
- OpenAI Academy: Small-business events
Workflow scoping review
Scope the smallest useful version of one workflow
Have one repeated task in mind? BaristaLabs can help define the minimum sources, the employee review point, the actions that stay blocked, and the evidence a first pilot should produce.
Best fit for teams that already have one candidate workflow and a named owner, but need help setting the boundary for a first pilot.
Self-serve readiness check
Check one workflow before you book a review
Answer nine questions about one workflow. You’ll receive a rules-based result covering impact, effort, risk, ownership, and the next step, with no booking required.
- 3–5 minutes
- 9 questions
- Use broad descriptions only
Best fit for teams that want to assess a workflow idea on their own before deciding whether to revise it or discuss a pilot.
Practical AI Workflow Notes
Want more practical AI operations ideas?
Get short notes on applying AI inside real small-business workflows — from document handling and customer follow-up to internal reporting, compliance, and automation guardrails.