Fifty-seven percent of U.S. small businesses are now investing in AI — up from 36% in 2023, according to research from Business.com. Only 14% say AI is fully embedded in their core operations. That gap is not a technology problem. It is a leadership problem.
The tools exist. The pricing has come down. The use cases are documented across every industry. What most small and mid-market businesses are missing is not another AI subscription but clarity on the organizational questions that tool selection cannot answer: What do employees do differently with AI than without it? How does AI change how work is organized? Who owns the outcome when an AI-assisted decision goes wrong?
stateDiagram-v2 direction TB state "Tools acquired" as T state "Workflows still unchanged" as U state "Redesigned around AI" as R state "Embedded in operations" as E state "Measured and optimized" as M [*] --> T T --> U: "We have AI" — most stop here T --> R: Leadership answers the process questions R --> E E --> M M --> [*]: 37% productivity gain territory U --> R: Leadership intervention
The Sequence Most Organizations Get Wrong
The standard pattern is: identify a tool, buy a subscription, tell employees it is available, wait for productivity to improve. That sequence rarely produces the 37% productivity improvement that research describes — because the productivity gain comes from workflow redesign, not tool access.
Ethan Mollick, a Wharton professor who studies AI adoption across organizations, stated the problem plainly in a recent post: “A corporate position that workers should ‘just use AI to do stuff’ has never been enough. AI use in companies is a leadership problem that involves answering fundamental questions about what people should do with their time, how work is organized, and how to center people in work.”
That holds at every scale. A 12-person business telling employees to use an AI writing tool and a 12,000-person enterprise telling divisions to adopt a copilot are making the same error at different magnitudes. Subscriptions do not redesign work. Leaders do.
The Three Questions That Come Before Tool Selection
Before selecting any AI tool, an organization needs answers to three questions that only leadership can resolve.
What work is this replacing or augmenting? AI works well on high-repetition tasks where quality can be defined and measured. It works poorly as a general productivity concept. If you cannot describe which specific tasks will change and by how much, tool selection is premature.
Who owns quality in the new workflow? AI tools produce outputs — drafts, summaries, analyses, recommendations. Someone is still accountable for those outputs: the customer email that goes out, the report that reaches a client, the analysis that drives a decision. Embedding AI without clarifying accountability creates diffusion of responsibility that surfaces as quality problems downstream. Assigning that accountability is a leadership decision, not a technical one.
How will you know if it is working? The default measure of AI adoption is how many employees have accounts or how frequently the tool is opened. The useful measure is what changed in the underlying work — time per task, error rate, output volume, customer response quality. Without measurement before and after, there is no learning, and the investment runs blind.
What Embedded AI Actually Looks Like
The 14% of small businesses reporting AI as fully embedded in operations are not using more tools than the 57% who are investing. They are using fewer, more deliberately.
The pattern is consistent: one workflow gets redesigned around AI first. It is chosen because it is high-repetition, the output quality is measurable, and the team that owns it is willing to serve as the test case. The implementation is documented — what the old workflow looked like, what the new one looks like, what changed in output and time. That documentation becomes the template for the next workflow redesign.
This is slower and more deliberate than “give everyone access and let them experiment.” It is also the sequence that produces embedded AI rather than subscribed AI. Sixty-two percent of SMBs that have successfully embedded AI in customer service and marketing did so by being specific about what AI was doing, who was accountable for the output, and how they knew it was better than before.
The Expertise Barrier Is Real but Bounded
Lack of expertise is cited as a barrier by 54% of small businesses — second only to cost. That is a real constraint. Understanding which tools are worth the cost, how to integrate them into specific business workflows, and how to measure their impact requires either someone inside the organization who has done this before or access to that judgment from outside.
The value of fractional technology leadership in this context is not producing a recommended tool list. A list is available from any number of sources. What it provides is the organizational work: which workflows to redesign first, how to redesign them, how to measure the outcome, and how to apply those lessons to the next one. The sequence matters, and the sequence starts with questions the tools cannot answer.
Start with One Workflow
The most common cause of stalled AI adoption in small businesses is scope. Organizations that attempt to implement AI across every function simultaneously end up with subscriptions that are not being used and no evidence of whether any of them work.
The right scope is one workflow, defined specifically, with a measurable outcome. Get it working. Document what changed. Then move to the next one.
The 14% figure is not the ceiling. It is where organizations are that have run this process at least once. The remaining 43% who are investing in AI without embedding it are still in the “we have the tools” stage, waiting for productivity to arrive on its own. It does not arrive that way.
