Which Businesses Should Switch to AI Now — and Which AI Tools Fit Each Department (2026)

1. Where AI adoption really stands in 2026

Headlines often claim that "almost every company uses AI." Official statistics, which survey a representative sample of all businesses rather than mostly large ones, tell a more useful story: about one business in five uses AI to produce goods or services.

The gap between 20% and 88% is not a contradiction. McKinsey surveys executives at organizations already investing in technology and counts any regular use in one business function. Statistical offices ask every kind of business — including the corner shop and the five-person accounting firm — whether they use AI in producing goods or services. Even in McKinsey's sample, only about one-third of companies had begun to scale AI beyond pilots, and just 6% qualified as "high performers" attributing 5% or more of EBIT to AI.

Adoption is rising fast. In the EU the share more than doubled in two years, from 8.1% in 2023 to 20.0% in 2025. In the US, 20–23% of businesses expected to be using AI within six months in early 2026.

EU enterprises using AI, by year

Size is the strongest predictor

Large companies adopt AI three times as often as small ones. They have data teams, budgets, and processes that are already digital. Small firms are catching up, but the gap remains large.

EU enterprises using AI in 2025, by size

The US pattern is the same: in the period ending May 3, 2026, about 37% of US businesses with 250 or more employees and 32% of those with 100–249 employees used AI, while fewer than 20% of firms with four or fewer employees did.

Some industries are far ahead

US businesses using AI by sector (as of May 3, 2026)

In the EU, information and communication (62.5%) and professional, scientific and technical services (40.4%) lead as well. These sectors share three traits that matter for every business deciding whether to switch: their product is information, their work is mostly text and numbers, and their data is already digital.

What are firms using AI for? Among EU enterprises that use AI, the most common purposes in 2025 were marketing or sales (34.7%) and business administration or management (31.1%). The most used technology was text analysis (11.8% of all enterprises), followed by generating images, video, or audio (9.5%) and generating written or spoken language (8.8%).

2. What AI measurably changes — the evidence

Vendor case studies promise dramatic savings. Peer-reviewed research is more modest, and more useful, because it shows which tasks and which workers benefit.

Lesson 1: The biggest gains go to newer and less experienced staff

In the customer-support study, less experienced and lower-skilled agents improved in both speed and quality, while the most experienced agents saw small speed gains and small declines in quality. AI spreads the know-how of your best people to everyone else. That is valuable for businesses with high staff turnover or many new hires, and a warning not to let AI overrule your experts.

Lesson 2: AI works on tasks, not on whole jobs

Eloundou and colleagues, writing in Science, assessed AI exposure task by task rather than job by job. Most jobs contain some tasks that AI handles well (drafting, summarizing, classifying, extracting) and many it does not (negotiating, physical work, judgment calls with incomplete information). The right question is not "Can AI do this job?" but "Which tasks in this department are repetitive, language-based, and checkable?"

Lesson 3: Customer and employee experience can improve too

The same customer-support study found that customers were more polite and less likely to ask for a manager when agents had AI assistance. The writing experiment found that people exposed to ChatGPT were twice as likely to use it in their real job two weeks later. Once people see a tool save them time on a task they dislike, adoption tends to spread on its own.

3. Seven signs your business is ready — and should switch

Readiness research (Jöhnk, Weißert and Wyrtki, Business & Information Systems Engineering, 2021) groups the factors into strategy, resources, knowledge, culture, and data. In practice, these become seven observable signs. If four or more apply to you, waiting costs more than starting.

1. The same questions arrive every day. Price, stock, opening hours, order status, delivery times — answered by hand, again and again.
2. Someone retypes documents. Invoices, receipts, delivery notes, or forms are copied from paper or PDF into a spreadsheet or system.
3. Content is always late. Social posts, product descriptions, proposals, and reports pile up because they take hours to write.
4. Customers wait for replies. Messages sit unanswered after hours or during peaks, and you lose sales to faster competitors.
5. Your data is already digital. Sales, stock, and customer records live in a system or at least in organized spreadsheets, not on paper.
6. People already use ChatGPT on their own. Staff are experimenting without rules — a sign of demand, and of risk if customer data is pasted into personal accounts.
7. Leadership can name one measurable pain. "Replies take 6 hours," "month-end closing takes 5 days." A measurable pain is what makes a pilot provable.

4. Which types of business should move first

The businesses that benefit first are not necessarily tech companies. They are businesses where many small, repeated, language-heavy tasks pile up. The table summarizes where each type typically gets its first win.

Notice the pattern: in almost every case, the first win is a communication or document task, not a futuristic one. Forecasting, computer vision, and predictive maintenance pay off too, but they need clean historical data and usually a technical partner, so they come second.

5. When not to switch yet

Some businesses should hold off — or rather, do something else first. Use this quick check.

Two more red flags: if your team is in the middle of another major change (a new ERP, a merger, a reorganization), wait for it to settle. And if leaders expect AI to cut headcount immediately, reset that expectation — the evidence in section 2 shows gains in speed and quality with people still in the loop.

6. Which AI fits each department

"AI" is not one product. There are five families of tools, and each department needs a different mix:

• General AI assistants (ChatGPT, Claude, Gemini, Microsoft 365 Copilot) — drafting, summarizing, analyzing documents and spreadsheets.
• AI agents built into business software — customer-service bots, CRM assistants, accounting tools that read receipts.
• Automation platforms (Zapier, Make, n8n) — connecting apps so that AI steps run automatically.
• Creative AI — images, video, and voice for marketing.
• Predictive and vision AI — forecasting, anomaly detection, quality inspection; these need clean historical data.

Choosing a general AI assistant

For most small and mid-sized businesses, a business plan of one general assistant covers 60–80% of the first-year use cases. Choose the one that fits the software you already use: Microsoft 365 Copilot if your team lives in Word, Excel, and Outlook; Gemini if you run on Google Workspace; ChatGPT or Claude if you want a standalone assistant that works across tools. Whatever you choose, use a business or team plan rather than personal accounts, so company data is handled under business terms and access can be managed centrally.

7. What it costs to start

Prices below were checked on October 19, 2026. "Vendor" means the price was read on the vendor's pricing page; "reported" means the vendor page could not be read automatically and the price comes from several dated third-party price checks. Confirm on the vendor's site before buying.

A worked example

A 5-person team buys a general assistant at $25 per seat per month and one automation plan at about $20 per month: $145 per month. If the writing evidence holds for the drafting part of their work — 40% less time — and each person spends 5 hours a week drafting emails, posts, and reports, the team saves about 2 hours per person per week, or roughly 43 hours a month. At a loaded cost of $20 an hour, that is about $860 of time each month, around six times the subscription cost. Your own numbers will differ. That is exactly why section 9 starts by measuring the task before the pilot.

Want to add up an existing AI and software stack? Our AI Tool Cost Calculator totals subscriptions per year and flags overlaps.

8. The three real barriers and how to remove them

When EU enterprises that had considered AI were asked why they did not use it, three answers dominated. None of them is about the technology being too weak.

Why EU enterprises that considered AI did not use it (2025)

Barrier 1 — Expertise (70.9%)

You do not need a data scientist to start. You need two or three people who are trained on your chosen assistant with your own tasks. A two-hour hands-on session using real company documents works far better than a generic course. Appoint one "AI champion" per department to collect prompts that work and share them.

Barrier 2 — Legal uncertainty (52.5%)

Write a one-page AI use policy: which tools are approved, what data may never be entered (customer IDs, health data, salaries, unreleased financials, contracts under NDA), that every AI output used externally is checked by a person, and who to ask when unsure. In regulated sectors such as health and finance, have your legal adviser review it once.

Barrier 3 — Privacy (48.8%)

Use business plans with admin controls, turn off training on your data where the vendor offers that setting, and keep sensitive work inside systems you already trust. For customer-facing bots, publish what the bot does and offer a path to a human.

9. A 90-day adoption plan

The order matters. Automating a process before people trust the AI step usually automates mistakes. Manual first, then automatic.

10. Map your own company in 5 minutes

Which of your departments should start with AI?Free · 5 minutes · works on your phone
Start the AI Adoption Map →

11. Frequently asked questions

12. Sources

• US Census Bureau — "Large Firms With at Least 20 Employees Biggest AI Users", May 26, 2026 (Business Trends and Outlook Survey, data Dec 14, 2025 – May 3, 2026).
• Eurostat — "20% of EU enterprises use AI technologies", December 11, 2025; and "Use of artificial intelligence in enterprises", Statistics Explained.
• McKinsey & Company — "The state of AI in 2025: Agents, innovation, and transformation", November 2025.
• E. Brynjolfsson, D. Li, L. Raymond, "Generative AI at Work," The Quarterly Journal of Economics (2025), https://doi.org/10.1093/qje/qjae044.
• S. Noy, W. Zhang, "Experimental evidence on the productivity effects of generative artificial intelligence," Science 381 (2023), https://doi.org/10.1126/science.adh2586.
• T. Eloundou, S. Manning, P. Mishkin, D. Rock, "GPTs are GPTs: Labor market impact potential of LLMs," Science 384 (2024) 1306–1308, https://doi.org/10.1126/science.adj0998.
• J. Jöhnk, M. Weißert, K. Wyrtki, "Ready or Not, AI Comes — An Interview Study of Organizational AI Readiness Factors," Business & Information Systems Engineering 63 (2021) 5–20, https://doi.org/10.1007/s12599-020-00676-7.
• Pricing pages: Claude, Intercom, Zapier (read on October 19, 2026); ChatGPT and Microsoft 365 Copilot (reported prices, confirm on the vendor page).