Goldman Sachs committed $500 million on Thursday to launch a Small Business AI Institute at its lower-Manhattan headquarters, and in the same press packet published a survey showing why the money is necessary: 86% of graduates from the bank’s 15-year-old 10,000 Small Businesses program say they use AI, but only 15% have fully integrated it into core operations.
That’s a 71-point gap between awareness and results, and it’s the actual story.
The Babson College survey of more than 2,400 program alumni, conducted in August 2026, found AI usage up from 56% in 2024. More than three-quarters of respondents said they could use additional training and resources. An earlier Goldman survey of 1,256 participants, conducted January 27 through February 4, 2026 and published March 17 by Babson and David Binder Research, had already flagged the same pattern: 73% wanted more training and implementation help, and 87% said AI augments rather than replaces employees.
The Institute’s response is more training. Anthropic, Google, Meta, and OpenAI are the launch partners. The programming includes the Grow with Google AI Professional certificate for nearly 20,000 owners and employees, three months of no-cost access to Google AI Pro, and Meta’s Muse for Small Business. Three pillars: AI education, published research, and AI-enabled expansions to capital access. Asahi Pompey, Goldman’s global head of Corporate Engagement, framed the Institute as something “that is practical, accessible and immediate.”
CEO David Solomon, speaking at the launch, offered an aside that reads differently in context: “there’s very little that we do where we need frontier models.” Goldman’s own house view, applied to its own operations, is that the model layer isn’t the bottleneck. Yet the fix prescribed to small businesses is more access to that same model layer, from four of its largest vendors.
Mike DiMarino, who runs Brooklyn’s Linda Tool and was in the inaugural 2010 cohort, is the kind of operator the program has long celebrated; graduates collectively represent more than $32 billion in annual revenue and over 350,000 jobs, and add jobs four times faster than peers. Rechelle Balanzat, founder of the laundry app Juliette, told NBC she “really treat[s] my AI as my consigliere.” A consigliere, notably, is someone you still have to brief.
That’s the structural tell. The survey confirms what a New York Fed study earlier this month already suggested, that only one in three AI-using small businesses has seen sales grow, and what reporting on agency spend cuts implied, that owners are moving work in-house faster than they’re learning to run it. The gap isn’t knowledge. It’s time, selection, prompting, operation. More workshops don’t change that math.
LemonLime, positioned at lemonlime.com as “the AI that gets you customers,” proposes the inverse: decide which sales and marketing work matters for the business, and deliver finished output. No workshop required. Whether that proposition holds for any given owner is a separate question. The Goldman data just made it the right question to ask.