OpenAI’s SBDC partnership puts the follow-up after AI training in focus

OpenAI announced a partnership with America’s SBDC on September 30 to expand hands-on AI training and local guidance for small businesses. The initial plan is to train around 150 advisors and reach at least 1,000 businesses through in-person workshops. For an owner, the most useful part may be what happens after the class: SBDC advisors are expected to follow up as participants try the tools in their own work.

That is a plan for learning and support, not evidence that the participating businesses have already saved time or grown revenue. OpenAI released usage statistics alongside the partnership, but those statistics answer a different question. A business considering a workshop should keep access, adoption, and results separate.

What the partnership plans to provide

America’s SBDC plans to select advisors for an OpenAI Academy training and credentialing pathway. Advisors and staff will also prepare to run group sessions through a separate facilitator pathway and small-business workshop module. OpenAI says the partnership combines its tools with the advisors’ existing local business relationships.

The planned Academy Workshops are modeled on a three-hour, hands-on format. Owners would build and test ways to use ChatGPT on real business tasks and leave with a plan for putting them into practice. The announcement calls for at least one workshop at each participating SBDC network, followed by advising to work through questions and see what is useful.

The partners also plan practical guides and an industry-focused companion guide. Marketing, operations, manufacturing, exporting, and government contracting are potential topics, not a confirmed catalog of guides already available. Workshop and advising experience is intended to shape those resources.

OpenAI does not give a nationwide workshop calendar, identify all participating local networks, or specify an enrollment fee in this announcement. Nor does it promise a particular ChatGPT subscription to every attendee. Ask your local center whether it is participating, what dates and eligibility apply, and what account or materials the session requires. The America’s SBDC website is a starting point for finding the network; this article is not an enrollment notice.

Why follow-up advising changes the training decision

The July ChatGPT small-business program already brought together training, events, guides, and partner resources. September’s announcement adds a specific local-advisor channel with planned workshop follow-up. The new decision is not simply which ChatGPT plan to buy. It is whether a local learning opportunity can help you work through an actual task after the demonstration ends.

A class can show how to ask for a draft or compare a set of documents. Applying that technique at work raises more specific questions: which examples represent the task, which answers need correction, and whether the person doing the work still finds the approach useful next week. An advisor who can return to those questions offers something a recorded demonstration cannot.

Before attending, choose one recurring task and describe how it works today. Bring non-sensitive examples, the expected output, and the question you want help answering. For example, an owner testing a draft customer response could ask whether the draft preserves the facts and reduces total preparation time once editing is included. That is a proposed test, not a result reported by the program.

Our guide to what a two-day AI agent workshop can prove makes the broader distinction: operating a bounded practice workflow is useful evidence of learning, but it is not proof of sustained reliability or return on investment. The SBDC workshop’s planned three-hour format is a different engagement; the same distinction between practice and business evidence still matters.

BaristaLabs diagram: access and activity do not establish business results; those require records from the actual task.

OpenAI’s usage figures are not workshop outcomes

OpenAI says around 4 million employees of companies with fewer than 500 people used its tools worldwide during September 9–15. Nearly one in five of those users worked at a company with fewer than 10 employees. These are vendor-reported usage figures for a recent week, not the number of SBDC participants, a count of U.S. businesses, or a measure of productivity.

The announcement also says agentic output tokens accounted for two-thirds of small-business output tokens in August, up from one-third in April 2026. Tokens are units processed and generated by models. Their share describes usage composition; it does not establish that two-thirds of work was automated, that tasks were completed successfully, or that labor costs fell.

OpenAI includes owner stories about work such as comparing restaurant menu prices and preparing outreach. Those accounts can suggest exercises worth trying. They are not a controlled evaluation of this newly announced training pilot, and an individual reported sale does not establish the expected return for another participant.

Bring evidence back to the advisor

Use the follow-up to examine what happened on your task, not whether the training felt impressive. Record the time spent preparing inputs, producing the output, and checking or correcting it. Compare that total with how your team handles a similar task without AI. Keep examples of both useful outputs and ones you rejected, so the advisor can help distinguish a prompting problem from a task that is a poor fit.

A short follow-up note can answer four questions:

• What did we try? Name the task and the examples used, rather than “AI for operations.”
• What changed? Record preparation and review effort, errors caught, and work actually completed.
• What remains unresolved? List the cases that still needed the old method or specialist help.
• What should happen next? Continue practicing, revise the task, or stop using the approach for that work.

This is BaristaLabs’ suggested way to use advising, not a published SBDC reporting requirement. It makes the conversation concrete without demanding a large deployment or an elaborate measurement system. If the exercise warrants a longer trial, our guide to what a first AI pilot leaves behind explains how to preserve the evidence needed for the next decision.

The partnership’s promise is more accessible instruction backed by people owners can return to. Its business value will depend on what participants continue using and what improves in their actual work. Ask about local access now; reserve claims about results for the follow-up records.

Source

• OpenAI, September 30, 2026: Helping small businesses put AI to work. Partnership plans and vendor-reported statistics are attributed to this announcement.