76% of Small Businesses Use AI. Only 14% Have Built Anything Around It.

AI adoption among small businesses has moved faster than almost anyone predicted. According to a Goldman Sachs 10,000 Small Businesses Voices survey of 1,256 small business owners, 76% currently use AI, and 93% of those who do use AI report a positive impact on their business.

But there’s another number in the Goldman Sachs data that’ tells a very different story’s equally interesting. Only 14% of small businesses have AI fully integrated into their core operations. The rest are still in the early stages, experimenting with tools on individual tasks without changing how the business actually runs.

Most of the unrealized value exists somewhere between using AI and running on it. And it turns out the fix is more about planning and sequencing than buying better tools.

Adoption Is Complete. Integration Is the Real Work.

Among small business owners using AI, 84% cite increased efficiency and productivity as the primary benefit. And 87% say AI is augmenting their workforce rather than displacing employees. The satisfaction numbers are unusually strong for any technology this early in its lifecycle.

So the tools are working, people like them, and the results at the individual level are real.

What’s missing is the step that turns individual tool use into a business capability. Using AI means someone on your team opens ChatGPT or another tool when a task calls for it. Integration means a process was redesigned so the tool is part of how the work gets done by default, regardless of who’s doing it. The first is personal productivity, while the second is something the business can measure, repeat, and build on.

This Is an Enterprise Problem, Too

If you’re running a small or midsize business and feeling behind the large companies on AI, the data suggests you shouldn’t. The same disconnect between individual results and business-level impact shows up at every company size.

McKinsey’s 2026 State of AI survey found that 80% of respondents say AI has improved their individual productivity. Meanwhile, only 37% report any positive contribution to their organization’s EBIT, a figure that is essentially unchanged from the year before. Nearly nine in ten organizations are using AI regularly in at least one business function, and 44% report scaling it across the enterprise, up from 38%. But the financial needle hasn’t moved for most of them.

McKinsey classifies about 6% of respondents as “AI high performers”, defined as organizations that attribute at least 5% of EBIT to AI and describe its impact as significant. That share has been flat for two years running.

Where Most Companies Stand

So if 76% of small businesses are using AI and only 14% have integrated it, where is everyone else? A 2026 study from SAS and IDC, surveying more than 1,600 SMB leaders across 28 countries, mapped it out across four stages of maturity.

Experimental (37% of SMBs)

This is where individuals are using AI tools on their own initiative. There’s no shared account structure, no agreed-upon use cases, and no record of what’s working. The value is real, but it’s invisible to the organization, and it’s non-transferable. If the person using the tool leaves, there’s no clear process for someone else to continue what they’ve been doing.

Opportunistic (33% of SMBs)

Some deliberate use cases exist here, often in marketing or customer service. Tools were chosen with intent rather than stumbled into. But the AI efforts are still disconnected from each other and from the systems that run the business. Someone decided to use an AI writing tool for social media posts. Someone else is using a chatbot for customer inquiries. Neither initiative knows about the other.

Structured (22% of SMBs)

At this stage, use cases are chosen against business priorities. Someone owns the AI work. Data has started to be organized around it, and results are measured against something concrete. This is where the shift from “we use AI” to “we have an AI strategy” actually begins.

Integrated (9% of SMBs)

AI is part of how decisions get made and how work moves through the organization. Workflows were redesigned around it. This is the stage where the McKinsey high-performer data and the Goldman Sachs integration data point at the same set of behaviors.

What the Top 6% Do Differently

The McKinsey data on high performers is the most useful part of this entire conversation, because it shows that the differentiators are behavioral. They’re decisions, not purchases.

Nearly three-quarters of AI high performers report fundamentally redesigning workflows because of AI, up from 55% the year before. Among everyone else, that number is one-quarter. High performers are also 3.3 times more likely to intend to use AI to fundamentally transform the business within the next three years. They’re twice as likely to report visible leadership commitment and twice as likely to have defined processes for measuring AI impact.

And here’s an important detail that gets overlooked. About 80% of both high performers and everyone else set efficiency as an objective for their AI initiatives. Efficiency alone isn’t the differentiator. High performers set growth or innovation as additional objectives alongside efficiency. They’re asking “what can we do differently?” on top of “how do we do existing things faster?”

Strategy Starts With Boring Things

If the data makes one thing clear, it’s that the first strategy work is unglamorous. It involves governance, ownership, and measurement.

From the Founder Reports AI in the Workplace survey of more than 2,000 U.S. workers, 44% say their employer has no clear AI policy, or they aren’t sure if one exists. At companies with fewer than 10 employees, that number rises to 59%.

The Goldman Sachs data reinforces this from the other direction. Seventy-three percent of small businesses say more training and resources would help them implement AI successfully, and about half cite data privacy and security concerns as a barrier. The SAS/IDC study identified the same recurring barriers to maturity at the organizational level, including fragmented data and tools, isolated AI initiatives, limited skills, and insufficient governance.

Every one of those barriers is solvable by making a decision rather than buying something. A one-page AI usage guideline, a named person who owns AI decisions, and a baseline measurement on one workflow will put a small business ahead of the vast majority of its peers.

Building a Roadmap That Survives Contact With Your Business

The companies that reach production with their AI initiatives tend to share a few practical habits. None of them are expensive.

Pick a narrow first project. The best starting point is a task that happens frequently, can be measured, and has low consequences if something goes wrong. Responding to common customer inquiries, generating first drafts of proposals, or summarizing meeting notes are all reasonable candidates. The goal of the first project is to build the muscle for the second one, so choose something where learning is cheap.

Establish a baseline before you change anything. The most common reason an AI initiative can’t demonstrate value is that nobody measured the workflow before it was changed. If you’re going to use AI to speed up your quoting process, record how long quotes take now. If you’re going to use it for customer service responses, measure your current response time and resolution rate. You can’t report a result against a number you never captured.

Sequence data work before ambitious builds. Fragmented data is the most commonly cited structural barrier in the SAS/IDC study. If your business data lives in disconnected spreadsheets, inboxes, and individual employees’ heads, the AI tool will only be as useful as whatever someone manually feeds into it. Before planning a custom build or a complex integration, get your existing data organized and accessible.

Name one owner. Even in a five-person company, someone needs to own AI decisions. That means deciding which tools to try, reviewing whether they’re working, and being the person who says “we’re going to stop doing it this way and start doing it that way.” Without ownership, AI stays in the experimental stage indefinitely.

Set a review date. Decide in advance when you’ll evaluate your first project and what would make you expand it, change it, or stop. A 90-day review is reasonable for most small businesses. Having a date on the calendar prevents the two most common failure modes: abandoning something too early because results aren’t instant, and continuing something indefinitely because nobody asked whether it was working.

Looking Forward

AI operating costs are starting to show up as a real consideration. About one in five organizations in McKinsey’s survey report that AI-related costs, including token costs, have constrained their use. At the same time, 60% of organizations expect to increase AI investment in the coming year.

The build-versus-buy equation is shifting, too. Thirty-two percent of organizations say they’ve decided against purchasing at least one software product because they could build the functionality in-house using agentic coding tools. That number is likely to grow. For small businesses, this means the cost of building custom capability around AI is falling, which makes knowing what you want to build more valuable than it used to be.

Frequently Asked Questions