AI demand in numbers: what the adoption data means for business owners

Reported AI use rose across three survey years

The chart on this site uses the share of respondents saying their organization uses AI in at least one business function. Each point belongs to the year the survey took place. The 2024 result appeared in a report published in March 2025, so labeling it as a 2025 observation would shift the series incorrectly.^[2]

The change from 55% to 88% is 33 percentage points. It does not mean that 33% more businesses sold data. Stanford's 2026 AI Index also reports this adoption trend, providing a useful research reference alongside the original surveys.^[4]

Read the survey definition before using the percentage

These are weighted global surveys of respondents across industries and organization sizes. McKinsey adjusts responses using each respondent's country's share of global GDP.^[1][2][3] The figures are not a census of Orlando companies or a measurement limited to small businesses. In the 2025 sample, 38% of respondents worked at organizations with more than $1 billion in annual revenue.^[3]

The wording also deserves attention. The 2024 survey left adoption undefined, allowing responses to cover anything from a few employees experimenting to extensive deployment. The 2025 report describes regular use in at least one function.^[2][3] The series shows a broad direction across annual surveys; it does not follow an identical set of small businesses under a fixed operational test.

For an owner considering a license, a useful interpretation is that AI has become part of more organizations' work. North Fox's inference is that this makes practical business tasks worth investigating as a data category. It supplies no estimate of the chance that a buyer will accept your records.

The 2026 survey adds context about deployment

McKinsey's August 2026 report describes regular AI use as nearly nine in ten respondents and reports that 44% say AI is scaling across their enterprise. That is a different measure from using AI somewhere in the organization. Its fieldwork ran from May 4 to June 8, 2026, with 1,719 participants.^[5]

Our adoption chart ends with the precise 2025 annual result. We do not turn the newer report's rounded description into an invented fourth percentage, or append its scaling figure as though the definition were unchanged. Wider deployment may give developers more practical problems to solve. It does not identify which outside datasets they will purchase to solve them.

Buyers are asking about records of completed work

There is a more direct connection to licensing than adoption alone. Mercor's data program names business records such as work orders, quality logs, shipping documents, dispatch records, and project files. It also describes connected sources including email, workplace conversations, CRM records, and code repositories.^[6] Troveo's platform announcement includes enterprise workflow traces among its data categories.^[7] These are statements of program interest from the providers.

Consider an illustrative distribution workflow: an order arrives, a stock shortage appears, a dispatcher considers alternatives, a manager approves a substitution, and the customer accepts the delivery. A connected record can explain the problem, available information, decision, and result. A list of final order totals leaves much of that sequence out. This example explains what to inventory; it is not a claim that a buyer has requested or priced that particular collection.

Ask a prospective reviewer which part of the work would be useful: the decisions, the source documents, the final outcome, or a combination. Their answer should help narrow the proposed scope before anyone discusses access to systems.

Keep adoption, investment, and licensing evidence separate

A chart becomes less useful when its label promises more than its data. We use different evidence for different questions. An adoption percentage can describe use. A funding total can describe investment. Neither establishes the amount paid to a record owner. The table below is a reading guide, not a forecast.

Use the evidence to choose one collection to investigate

Start with a completed business process you understand well. For a contractor, that might be estimate through inspection. For a property manager, it might be a maintenance request through resolution. Choose a process with a clear beginning and end, then identify the systems that preserve its history.

Write a short description covering the work, approximate years available, related tools, and the person who understands the records. Separate facts you know from assumptions. Five years since installing a tool does not establish five years of complete, retained, usable history. A system migration or missing attachments may change the collection substantially.

Keep this first inventory descriptive. North Fox's review does not need customer names, actual messages, account access, or exported records. An Orlando-area owner can discuss the process in person and work out which questions need an answer before an approved introduction.

Ask five questions about a market claim

Before using a statistic to set expectations inside your business, trace it to its original source and check the following points. A clear answer is more useful than a larger headline number.

1. What exactly was measured: adoption, investment, advertised potential, an offer, or a completed payment?
2. Which year and period does it cover? Distinguish the survey date, publication date, transaction date, and the date somebody checked the page.
3. Who is included? Check organization size, geography, data category, and whether the examples were selected for marketing.
4. Which terms are public? Duration, exclusivity, deductions, delivery obligations, and record volume can change the meaning of a price.
5. What remains unknown? Preserve missing information instead of filling it with a confident assumption.

Move from broad evidence to a specific conversation

The public sources reviewed for this guide do not establish a representative annual spending series for AI licenses purchased from small businesses. Publishing one would require comparable transactions, clear seller definitions, known coverage, and consistent payment periods. Selected case studies cannot supply that dataset.

Use the adoption chart as context, then read the disclosed deal examples and the calculator methodology for their separate evidence. For your own organization, prepare the short inventory and identify who can approve a discussion. The next useful question is whether a current buyer wants that defined collection on terms your organization would accept.