Companies Using AI in Every Task: What the Surveys Show

Companies are not using AI in every task yet. Surveys of larger organisations report that about four in five use AI somewhere, government surveys of all businesses find about one in five, and in both the use clusters in a few functions: sales and marketing, writing and document work, IT and customer service. The financial gains companies report are real, and mostly small.

The original version of this post promised a future where AI runs every process. This version sticks to what companies report doing now, from four sources you can open: Stanford's AI Index, the US Census Bureau's business survey and a Census research paper, and Eurostat's survey of European enterprises. Each number below links to the page that published it.

The headline figure most people quote comes from a McKinsey survey charted in the 2025 AI Index. In 2024 78% of respondents said their organisation used AI, up from 55% in 2023, and the share using generative AI in at least one business function more than doubled, from 33% to 71%.

Government statistics, which sample businesses of every size, find much lower numbers. The US Census Bureau's Business Trends and Outlook Survey shows that AI use hovered between 17% and 20% of businesses from December 2025 to May 2026, with 37% of firms of at least 250 employees using it and fewer than 20% of firms with four or fewer. Eurostat found that 19.95% of EU enterprises with ten or more employees used AI in 2025, rising to 55.03% among large enterprises and falling to 17% among small ones.

The gap is mostly about who is counted. A survey of managers at large organisations asks whether AI is used anywhere in a big company, and the answer is almost always yes. A national survey asks every kind of business, and most businesses are small. If you run a company with twenty people, the second number is the one that describes your competitors.

The national numbers are moving, though slowly. In the same Census survey, between 20% and 23% of businesses expected to be using AI within six months, and the sectors ahead of the average were Information at 39.7% and Finance and Insurance at 33.9%, while Retail Trade sat around 14%. Use rose among firms with at least 20 employees and did not change significantly among smaller ones.

When companies do adopt AI, they start in the same places. A Census Bureau working paper found that, among firms using AI, the leading business functions are sales and marketing (52%), strategy and business development (45%) and IT (41%), and that writing, document analysis and information search are the leading generative AI tasks. Across all firms, it measured 18% using AI in a business function, or 32% when weighted by employment.

European data points the same way. Of the EU enterprises using AI, 34.70% used it for marketing or sales and 31.05% for business administration or management, while logistics came last. The pattern makes sense: these are the tasks with the most text, the most repetition and the clearest way to check the output.

That is also where AI agents, as opposed to chat tools an employee types into, fit first. Answering enquiries, booking appointments, following up on quotes and handling documents are high volume tasks with written rules, which is why our AI agents for business posts keep returning to them, from after hours WhatsApp bookings to the industry versions for clinics, brokerages and financial firms.

Companies that use AI do report results, but the size is smaller than the old version of this post implied. In the AI Index summary of the McKinsey data, 49% of respondents using AI in service operations reported cost savings, and most of those savings were under 10%. On revenue, 71% of those using AI in marketing and sales reported gains, with less than 5% the most common size of increase.

Jobs tell a similar story. The Census working paper found that most firms use AI to help people with tasks rather than to replace them, and that AI related employment decreases occurred in only 2% of firms. A small saving across many tasks still adds up, and it is usually how a company ends up using AI in more of its work: one task at a time, each one measured.

Eurostat also asked the enterprises that had considered AI but not adopted it why not. The top reasons were a lack of relevant expertise (70.89%), unclear legal consequences (52.52%) and data protection and privacy concerns (48.83%), while only 20.68% said AI was not useful for the enterprise.

Those are answerable concerns. The legal one in particular gets clearer once you look at the rules for a specific task rather than for AI in general: a clinic booking agent falls under HIPAA's business associate rules, a brokerage's AI callbacks under consent rules, a financial firm's client service agent under the conduct rules it already follows. The industry posts linked above go through those one by one.

Companies that end up using AI widely tend to add it task by task. The order below works because each step produces a number you can check before taking the next one.

1. Pick one task with volume and written rules: answering enquiries, booking, follow up or document handling.
2. Measure the baseline first, such as response time, missed calls, hours spent or error rate.
3. Write the scope and the handoff rules: what the agent does alone and what always goes to a person.
4. Connect the agent to the systems the task already uses, so it acts in your calendar, CRM or ledger rather than beside them.
5. Pilot on one channel, read the transcripts, and compare the result with the baseline.
6. Only then move to the next task, starting again at step one.

For a front desk, the numbers to compare are in AI voice agent vs a human receptionist, and the handoff rules in when an AI agent should hand off to a human. For work that does not fit a template, such as approvals, reconciliation or document processing, GenovaFlow builds custom agentic systems around the process you already run.