AI in business 2026: what 88% of companies already do and what it means for you

Back to blog

In McKinsey's latest The State of AI report, 88% of companies now use AI in at least one business function. [McKinsey 2025] In the previous report, published in July 2024, the figure was 78%, and the year before that, 55%. The growth has been faster than in any previous technology shift.

Where is AI already at work?

Customer service is one of the most natural and quickly measurable use cases. The reason is logical: customer service is repetitive, measurable, and largely predictable. The same questions come up, the same information gets collected, the same processes run day after day.

Klarna, the Nordic payments company, published the results of its AI agent rollout in customer service in 2024. [Klarna 2024] Resolution time dropped from 11 minutes to 2 minutes — 82% faster. Customer satisfaction stayed at the same level as human service.

According to Zendesk's analysis, the ROI of AI in customer service is $3.50 for every dollar invested. [Zendesk 2026] The figure is based on Zendesk's survey of customer service leaders. It describes a typical result, not a guarantee.

Market growth shows the direction

The conversational AI market was worth $12 billion in 2024. It's forecast to grow to $47 billion by 2030. [Fortune Business Insights 2024] The figure covers chatbots, voice agents, and other forms of AI-based interaction broadly — not just phone service.

This isn't speculation. It's a reaction to the fact that the technology works, and companies are seeing the results.

What's worrying, and is it justified?

The concerns are real. KPMG's 2025 global study found that only 62% of business leaders believe their organization is adopting AI responsibly. [KPMG 2025] Among employees, the figure is just 55%.

The biggest concerns:

• Cybersecurity risks: 51% of respondents
• Inaccuracies: 50%
• Privacy: 43%
• Workforce impact: 35%

These concerns are well-founded. According to McKinsey, 51% of organizations using AI have experienced at least one negative consequence. [McKinsey 2025] Inaccuracies are a real risk if a system hasn't been designed and tested carefully enough.

But fear doesn't mean waiting is safer. Your competitor is already making their decision.

The practical lesson: start with one narrow task

The companies that get the best results don't try to AI-enable everything at once. They pick one clear, repetitive task, measure the results, and expand from there.

Phone service is one of the most natural starting points. The task is concrete: answer calls, collect the essential information, and route it to the right person. AI does it nights and weekends, with no sick days or stress during busy periods.

AI for small businesses: how to start without a big AI project

You don't need a project team, a consultant, or a special budget. A model that works in practice has three steps:

1. Pick one repetitive task. What comes up day after day? Answering calls, managing bookings, order confirmations?
2. Measure where you stand now. How many cases per week? How much goes unhandled because of busy periods or absences?
3. Trial it for 7 days, narrowly. One task, one tool, one clear metric. Results show up fast. If they don't, you'll know without a big expense.

Adopting AI doesn't mean an organizational overhaul. It can be a week-long trial.

Example: property maintenance

A hypothetical example of a typical situation. An HVAC company received around 40 calls a week. Most were fault reports and urgency assessments. The owner was often out in the field, so a third of calls went unanswered.

A voice agent was set up. It answered calls, assessed urgency based on the customer's description, and sent a summary by email. In the first week, every call got answered. The owner called people back in priority order.

The change didn't require new systems or equipment. The only step was activating call forwarding from the phone.