The adoption gap is real—but it is not a reason to rush
Statistics Sweden reported that 16.1% of micro-enterprises with 0–9 employees used AI in 2025, up from 10.8% in 2024. Among enterprises with 10 or more employees, usage reached 35%; the figure was 30.8% for businesses with 10–49 employees and 71.9% for large enterprises. [1] [2]
Those figures describe adoption, not business value. A small owner-managed company should not copy the operating model of a 250-person organisation. Its advantage is a shorter decision path: one useful experiment can be chosen, observed and stopped without a transformation programme.
Start where the work is repetitive and reviewable
SCB found that marketing and sales and business administration or management were the two most common purposes among AI-using enterprises in 2025. These are broad categories, so the practical question is smaller: which repeated piece of preparation consumes attention but still has a human who can judge the result? [2]
Good first candidates include summarising a weekly operating record, preparing a first draft of a customer FAQ from approved information, comparing a website page against a fixed checklist, or organising cited competitor observations. Avoid starting with automatic pricing, employee assessment, legal conclusions or unsupervised customer communication.
Measure a business outcome, not the number of prompts
A prompt count is activity, not evidence of improvement. Choose one baseline that matters: minutes spent preparing the weekly review, percentage of recommendations accepted, number of factual corrections, or time from a detected issue to an owner decision.
If the workflow saves no time, creates repeated corrections or produces work nobody uses, stop it. A documented “not useful” result is better than allowing an unmeasured tool to become part of the business by habit.
USE THIS
A 30-day operating experiment
-
Name one outcome
Write the result in ordinary language and choose one baseline number.
-
Set the information boundary
List what the tool may use and remove unnecessary personal data.
-
Keep one human owner
The same person reviews factual accuracy, usefulness and risk.
-
Run a small weekly loop
Observe, prepare a draft, approve or reject it, then record what happened.
-
Decide after 30 days
Continue, change or stop based on the baseline—not on novelty.
PRIMARY SOURCES
Evidence used in this guide
-
Statistics Sweden (SCB)AI use in enterprises 2025
Micro-enterprise adoption and size-class comparison.
-
Statistics Sweden (SCB)Artificial intelligence in enterprises 2025
Adoption, barriers and common business purposes.