Executive summary: On October 22, 2026, Goldman Sachs published the third annual AI survey of its 10,000 Small Businesses program — 2,473 owners, fielded by Babson College — and committed $500 million to a new Small Business AI Institute built with Anthropic, Google, Meta and OpenAI. The survey's headline is both impressive and quietly alarming: 86% of respondents now use AI in their business, up from 56% in 2024 and 77% in 2025. But only 15% say AI is fully integrated into their core operations, and three out of four AI-using owners spend less than $100 a month. In other words, most small businesses are not AI-poor; they are AI-shallow. The gap is not a tool problem and it is not a budget problem. It is a sequencing problem: owners are pointing a general chatbot at everything instead of pointing the right tool at the few jobs that make money. This playbook gives four steps to close that gap this month, using the free training Goldman just put on the table.
The survey's most useful detail is where owners actually deploy AI. Financial analysis and business planning jumped from 18% in 2024 to 42% in two years; data analysis climbed from 26% to 42%; productivity and notes sit around 66%. But sales and customer-facing work — the stuff that closes the job — is the biggest untapped gap. Most of the 86% who use AI run it through a standalone chatbot (ChatGPT, Claude, or Gemini) and use it for writing and planning, not for the estimate, the invoice, or the follow-up that turns into revenue. So the first move is scope, not spend: name the single money job. For a service business it is usually the estimate or the follow-up. For a product business it is the order confirmation or the reorder prompt. One job. If you spread across five, you will measure none of them and integrate none of them.
Tools & pricing: A notebook. The decision costs nothing; the mistake is picking the fun job (marketing copy) over the money job (the invoice that gets paid).
Eighty-six percent of AI-using owners rely on a standalone chatbot as their main interface. That is fine for drafting an email, but a general window has no memory of your price sheet, your job types, your margins, or your customer history. The owners in the top 15% who have AI "fully integrated" are not using better models — they are using the same models pointed at a specific, repeatable task with the right context. So take your one money job and give a tool the context: your rates, your product or service menu, your typical terms, your past examples. Then have it draft the exact artifact you bill for — the estimate, the invoice, the follow-up message — and stop using the general chatbot for that job. This is the single highest-leverage change in the survey, and it costs the same as the chatbot you already pay for.
Tools & pricing: The AI tool you already run (a standalone assistant, or an in-suite agent like QuickBooks' AI agents or Microsoft 365 Copilot, both named in Goldman's report). No new purchase required to start.
This is the "why now." The $500 million Institute is not a vague pledge. As part of its initial activities, Goldman is partnering with Google to deliver its "Grow with Google" AI Professional certificate to almost 20,000 small business owners and their employees, including three months of no-cost access to Google AI Pro. It is developing a learning series with Anthropic on applied AI for small businesses, running hands-on workshops for 10,000 Small Businesses alumni, and partnering with OpenAI to deliver tailored AI workshops across the country with community-college partners. Meta is helping support owners with its Muse for Small Business. For an owner who has been typing into a chatbot alone, this is a free curriculum that would otherwise cost money: structured, vendor-backed, and aimed specifically at applying AI to a real business rather than at passing an exam. You do not need to be a Goldman program alumnus to learn from the same material — the underlying "Grow with Google" certificate and the public workshops are the starting point.
Tools & pricing: Google "Grow with Google" AI Professional certificate (free); three months of Google AI Pro at no cost for Institute participants; Anthropic applied-AI learning series; Meta Muse for Small Business (free for most uses). Verify current availability and eligibility directly with each provider.
Three out of four AI-using owners in the survey spend less than $100 a month, and 63% use AI every day. That means the constraint is not cost — it is trust and control. The right owners will want an agent that drafts the estimate but does not send it without a look, that prepares the invoice but does not send the reminder on its own, and that never moves money without a click. Set that rule before you let the tool run, not after you have to catch it. Every current agent — Duet, Dots, Muse, or an in-suite agent — exposes some approval gate for sending, paying, and signing; the owners who cross the 86-15 gap are the ones who configure it deliberately. Then measure the one job for two weeks: how long it took by hand, how long it takes with the tool drafting, and how many of those artifacts actually convert to revenue. That one number — not the tool's logo — decides whether you scale it.
Tools & pricing: A two-week measurement. The discipline, not the dashboard, is the deliverable.
The shift the Goldman data points to is not "adopt AI" — most owners already have. It is "sequence it." Eighty-six use it; fifteen run on it. The gap between those two is not closed by a new model, a bigger budget, or a fancier chatbot. It is closed by pointing the tool you already have at the one job that makes money, taking the free structured training now that the biggest banks in the country are paying for, and putting a human gate on the money step before you scale. The realistic outcome is not "10x revenue." It is the estimate, the invoice, and the follow-up moving from your desk to a tool — at the same near-zero cost you are already paying — so the jobs your competitors are still doing by hand are the ones you close first. That is how the fifteen end up buying the eighty-six's competitors, not the other way around.
Your next twenty minutes should be a scope decision, not a strategy debate. Write down the one job that makes you the most money. Pick the tool you already pay for and give it that job's context. Start the free "Grow with Google" AI certificate this week. Set the "human on the money" rule. Then measure for two weeks. To get the full SMB AI adoption strategy — including how this stacks with the other 2026 tools and where the free training actually pays for itself — start here:
Get the SMB AI adoption strategy