Goldman Sachs is putting $500 million behind AI training for small businesses, with Anthropic, Google, Meta and OpenAI as partners. The question it raises is whether training can close the gap between businesses that use AI and businesses that have actually changed how they operate. The survey the bank's coverage cites says that gap is still wide, which makes this a useful signal for solo operators, not just a press release.
What happened
On Thursday, October 22, 2026, Goldman Sachs launched its Small Business AI Institute at its headquarters in lower Manhattan. The launch sits inside a new $500 million commitment to the next chapter of the bank's 10,000 Small Businesses program, which covers education, access to capital and support services. Fortune reported the details from the bank's announcement, and NBC News covered the partner lineup. Nearly 20,000 entrepreneurs have completed the program since it launched in 2010.
Each partner brings a different piece. Anthropic is providing a learning series on applied AI for small businesses, hands-on workshops for program alumni and research on how small businesses actually use AI. Google is offering its Grow with Google AI Professional certificate to almost 20,000 small-business owners and employees, with three months of free access to Google AI Pro. Meta is supplying AI tools and training through what it calls Muse for Small Business. OpenAI is running AI workshops through the program's community college partners.
The question the story raises
Is AI actually reaching the businesses that need it most? The Babson College survey cited by Fortune suggests partly, but not in a way that has changed much. It found 86% of businesses now use AI, up from 56% in 2024, yet only 15% describe AI as fully integrated into core operations. Seventy-seven percent say they would benefit from more training and resources.
Among program graduates who use AI, 73% spend less than $100 a month on it. Most use chatbots such as ChatGPT, Claude or Gemini, and 18% use agentic AI. CEO David Solomon put the cost issue plainly: "At the moment, this is an expensive technology," he said, and "access to capital continues to be a constraint." He also made a point about Goldman's own work that applies to most small firms: "There's very little that we do where we need frontier models."
What it means for solo operators and marketers
First, the training is getting cheaper and more standardized. When four frontier labs compete to teach small businesses, the courses and certificates improve, and clients will start asking whether you hold them. Read the terms on Google's three-month offer before you build a plan around it, because promotional access has a way of expiring on a schedule nobody remembers.
Second, the gap between using AI and integrating AI is where paid work lives. Most operators have a chatbot subscription. Far fewer have an agent that pulls leads from a CRM, drafts follow-ups and writes results back into the system of record. That integration layer is the business. Buying another seat is not.
Third, the spend numbers argue for cost discipline. If nearly three-quarters of AI-using graduates spend under $100 a month, a solo operator running a dozen agents on premium model tiers is probably overpaying. Route routine tasks to smaller models and save the expensive ones for judgment calls.
The announcement also has limits. The reporting gives no timeline beyond the October 22 event, and a $500 million commitment is a headline figure, not a published spending schedule. Treat it as a signal of intent until the bank releases program-level details.
Our take
After 25 years in digital marketing, I have learned to be suspicious of any program whose logo wall is longer than its curriculum. Four AI labs on one stage is good theater, and I will say so with a straight face. Nothing says "the future of work" quite like four rivals posing for the same photo.
Still, I would rather a bank fund the training than hold the margin hostage, and the survey is the most honest part of the story. Most small businesses have adopted AI and few have changed how they operate. The tools are not the bottleneck. Workflow design is. Solomon's point that most work does not need a frontier model is worth framing above every AI budget meeting. Take the free certificate, skip the premium seat, and spend the savings on the integration nobody wants to pay for.