US businesses use generative AI in 2026 mainly to write, summarize, answer customers, create marketing content and help with code. A smaller group has built it into daily operations. Use is wide but shallow: 26 percent of US employees use AI at work a few times a week or more, according to Gallup. Most companies sit in the gap between trying generative AI for business and measuring a return.
The data below comes from 2025 and 2026 surveys by the Census Bureau, Gallup, McKinsey, the U.S. Chamber of Commerce and the Federal Reserve Bank of San Francisco. Each one counts something slightly different, and those differences explain most of the confusing headlines.
On this page
- How many US businesses use generative AI in 2026
- How US employees use AI at work
- The most common uses of generative AI for business
- How different US industries use generative AI
- Where generative AI produces measurable value
- How small businesses use it differently from large companies
- What generative AI costs a business in 2026
- The risks US businesses are managing
- How generative AI is changing how customers find you
- Where a US business should start
How many US businesses use generative AI in 2026
The Census Bureau gives the strictest count. Its Business Trends and Outlook Survey found 19.8 percent of US businesses used AI to produce goods or services as of May 3, 2026. About 20 to 23 percent expected to within six months.
Use varies sharply by sector. Information firms led at 39.7 percent and finance and insurance followed at 33.9 percent, while retail trade sat near 14 percent. Size matters too: 37 percent of firms with 250 or more employees use AI, against fewer than 20 percent of firms with four or fewer.
Broader surveys report higher numbers because they count any use, including an owner typing into ChatGPT. The U.S. Chamber of Commerce 2025 small business technology report found 58 percent of small businesses use generative AI, up from 40 percent in 2024 and 23 percent in 2023.
The Federal Reserve Bank of San Francisco sits in between. Its July 2026 brief on the Small Business Credit Survey reports that nearly 40 percent of small business respondents were using or planning to use AI.
Read together, the surveys say most small businesses have tried generative AI. About one in five has made it part of how they produce their product or service.
How US employees use AI at work
Employee data shows the same gap between trying and relying. Gallup's Q4 2025 workforce survey covered 22,368 employed US adults and was published in January 2026. It found 26 percent use AI at work a few times a week or more, and 12 percent use it daily. Another 49 percent never use it.
Only 38 percent said their organization has integrated AI to improve productivity, efficiency and quality. Another 41 percent said it has not, and 21 percent did not know. That last figure points to a communication problem as much as a technology one.
The most common tools are general assistants. Gallup reported in 2025 that about 6 in 10 employees who use AI rely on chatbots or virtual assistants. About 4 in 10 use it to consolidate information, generate ideas or learn new things.
Much of this use happens outside official tools. The KPMG and University of Melbourne 2025 Trust in AI study found 46 percent of US workers have uploaded sensitive company information to public AI platforms. Another 53 percent avoid disclosing when they used AI.
The most common uses of generative AI for business
The San Francisco Fed brief lists the uses small firms report: productivity tasks, marketing and SEO, written communications, visual design, customer service, analytics and custom tools. This table shows what each looks like in a typical US small business.
Use | What it looks like day to day | Risk to manage |
|---|---|---|
Writing and editing | Drafting emails, proposals, job posts and policies | Generic tone and wrong facts in client-facing text |
Marketing content | Social captions, ad variations, product descriptions, blog drafts | Claims that marketing cannot back up |
Customer service | Chat replies, email drafts, AI phone answering | Confident wrong answers to customers |
Summaries | Meeting notes, long email threads, contracts and reports | Missed details in the parts nobody reread |
Coding help | Writing scripts, fixing website bugs, building small tools | Insecure code and pasted credentials |
Data questions | Asking plain-English questions about sales or survey data | Wrong formulas presented as findings |
Search over your documents | An assistant that answers from your manuals and policies | Outdated or unapproved documents in the index |
Images and design | Mockups, social graphics, presentation visuals | Copyright and brand consistency |
Customer-facing uses carry the most risk and the most visible payoff. A phone assistant, for one, can answer after hours and book jobs, as covered in this guide to AI receptionists for small businesses.
How different US industries use generative AI
The same tools show up everywhere, but the first valuable use differs by industry. This table lists a strong first use for each and the limit that applies.
Industry | First high-value use | Limit to respect |
|---|---|---|
Home services | Answering calls and drafting quotes from job notes | Final prices stay with a person |
Professional services | Drafting letters, summarizing documents and research | A licensed professional reviews every client deliverable |
Healthcare practices | Front-desk scripts, internal policies and vendor paperwork | Patient data only with a vendor under a business associate agreement |
Retail and ecommerce | Product descriptions and answers to order questions | Accurate stock, shipping and return details |
Real estate | Listing descriptions and fast replies to new inquiries | Fair housing rules in every listing |
Manufacturing and distribution | Summarizing orders, specs and supplier emails | Quality and safety data checked against the source |
The pattern holds across rows. Generative AI drafts and summarizes, and a person signs off on anything that carries legal, financial or safety weight.
Where generative AI produces measurable value
The McKinsey State of AI survey, published in August 2026 with 1,719 participants in 97 countries, shows where returns appear. Respondents most often report revenue gains in marketing and sales, then in product development and software engineering. Cost benefits show up most in supply chain management, service operations and manufacturing.
Broad use is common, but bottom-line impact is not. In the same survey, 56 percent of organizations use AI in three or more functions, and 44 percent say they have scaled it across the enterprise. Only 37 percent report any impact on EBIT.
The difference is workflow design. Close to three-quarters of AI high performers report fundamentally redesigning workflows because of AI, against about one-quarter of other respondents. Giving every employee a chatbot does not change how work moves, while rebuilding one process around AI does.
That is the practical lesson for a small business. Pick one process, such as lead intake or quote follow-up, and redesign it with AI steps and a human check. The guide to AI workflow automation explains where to start.
How small businesses use it differently from large companies
Large companies build their own AI tools, while small companies buy them. McKinsey found 40 percent of organizations with more than $1 billion in revenue are scaling AI agents, against 22 percent of smaller organizations. Most small businesses use off-the-shelf assistants and the AI features already inside their CRM, email and design tools.
The San Francisco Fed brief names four barriers small firms report. They are the cost of the right tools, staff time for training and upgrades, knowledge gaps about implementation, and policy or regulatory limits. Firms that chose not to adopt cited accuracy and intellectual property concerns, the importance of human contact to their business, and a lack of relevance to their work.
Those reasons are sound. A family law practice or a custom furniture maker sells personal attention, and generative AI belongs in the back office there. A distributor answering 400 order-status emails a week has the opposite profile.
What generative AI costs a business in 2026
Generative AI is cheap to start and gets expensive only when you build. These are the main cost lines as of October 2026.
Option | Cost | Fits |
|---|---|---|
ChatGPT Business | $20 per user a month billed annually, $25 monthly, 2-seat minimum | Teams that need a safe general assistant |
Model API, mid tier | $2.00 per million input tokens and $10.00 per million output tokens for gpt-6.1-sol | Custom tools and automations with moderate volume |
Model API, small tier | $0.10 input and $0.50 output per million tokens for gpt-6-luna | High-volume sorting and labeling |
Custom assistant or integration | Priced by scope and developer time | Answers from your data or writes into your systems |
Sources: OpenAI business pricing and the OpenAI API price list. For custom builds, the cost drivers are broken down in the post on AI chatbot development cost.
Say a 15-person marketing agency in Atlanta moves everyone to ChatGPT Business on annual billing. That is 15 seats at $20, or $300 a month and $3,600 a year. If each person saves an assumed 2 hours a week on drafts and summaries, the team gains about 120 hours a month. At an assumed $40 loaded hourly cost, that time is worth about $4,800 a month.
The risks US businesses are managing
The main risk is unmanaged use, not the technology. IBM's 2025 Cost of a Data Breach report found one in five breached organizations had a breach linked to shadow AI, meaning staff using unapproved tools. High levels of shadow AI added an average of $670,000 to breach costs. Of breached organizations, 63 percent had no AI governance policy or were still writing one.
Four risks deserve a written rule in any US business:
- Data leakage, when staff paste client or employee data into personal accounts. This guide on whether ChatGPT is safe for business data compares plan terms.
- Wrong answers stated with confidence. The causes and fixes are covered in AI hallucinations in business apps.
- State AI laws on disclosure, hiring tools and automated decisions, summarized in AI laws that apply to US businesses in 2026.
- Customer trust, where AI-written replies or AI phone agents need clear disclosure.
This section is general information, not legal advice. A one-page AI use policy and a business-tier tool cover most of these risks for a small company.
How generative AI is changing how customers find you
Generative AI also changes the buyer side. Prospects now ask ChatGPT, Gemini and Google's AI answers for recommendations before they visit a website. A business that only ranks in classic search results can be missing from those answers.
Getting cited depends on clear, specific pages that AI systems can quote: prices, service areas, comparisons and direct answers to common questions. The approach is called generative engine optimization. Treat it as part of marketing, alongside SEO and ads.
Where a US business should start
Most businesses get value fastest by following this order:
- Approve one business-tier AI tool and move staff off personal accounts.
- Publish a one-page AI use policy covering data, review and disclosure.
- List the three writing or summarizing tasks your team repeats most each week, and time them.
- Measure time saved after 30 days, and drop any use that did not save at least an hour per person per week.
- Redesign one high-volume workflow with AI steps and a human approval.
- Consider a custom tool only when off-the-shelf assistants cannot reach your data or systems.
Step 6 is where the bigger gains sit for companies with their own software or large document sets. This guide covers adding AI features to an existing app or website.
Ask each team lead this week which task they would hand to an AI assistant first and how many hours it takes now. Those answers give you a ranked list of use cases and a baseline to measure against.
When a use case needs your own data or systems, the AI development team at Techno Mantra builds assistants, automations and LLM features on OpenAI, Claude or Gemini. You own all code and accounts, and an NDA is available on request. Send your use case through the contact page or to info@technomantraa.com. Every inquiry gets a reply within one business day.
Frequently asked questions
Generative AI creates new content, such as text, images or code, from a prompt. Traditional AI classifies or predicts from data, such as scoring a lead or flagging a fraudulent payment. Many business tools now combine both, using prediction to decide and generation to write the result.
Information and finance lead. The Census Bureau Business Trends and Outlook Survey found 39.7 percent of information firms and 33.9 percent of finance and insurance firms used AI to produce goods or services as of May 2026, against about 14 percent in retail trade.
Not at scale yet. In McKinsey's August 2026 State of AI survey, two-thirds of respondents reported little or no AI-related change in total employment over the past year, and 14 percent said AI contributed to a decline. Looking ahead, 39 percent expect AI to reduce head count over the next year.
About $40 a month for two people. ChatGPT Business costs $20 per user a month on annual billing as of October 2026, with a 2-seat minimum. API-based tools cost cents per thousand short tasks, and custom builds are priced by scope and developer time.
Many accept it if they are told. In an October 2025 OnePoll survey of 6,000 US adults for AnswerConnect, 86 percent said it should always be clear when they are speaking to AI. Disclose AI in customer conversations and keep a fast route to a person.