Irish small business owners have taken to AI. Nearly nine in ten SME leaders now use it at work, and on average they save themselves 5.3 hours a week, according to research Opinium carried out for OpenAI in March.
Their businesses haven't caught up. When the Small Firms Association surveyed 404 small firms in August, only 18% said AI was part of most or some of their business practices.
That's the gap in 2026. The owner and individual staff are getting hours back. The business, mostly, isn't. Business use is climbing, though: the CSO found AI use across Irish businesses went from around 8% in 2023 to just over 20% in 2025. Its 2026 figure isn't out yet.
And it isn't only small firms. McKinsey's State of AI 2026 survey found 80% of people say AI has improved their own productivity, yet only 37% say it has made any difference to their organisation's earnings, the same as last year.
So for a business of 10 to 50 people, the question isn't whether to use AI. It's how to move it from the owner's laptop into the way the business runs: where that pays off, where it doesn't yet, and how to start.
Where it pays off now
The questions you answer every day
Opening hours, prices, how to book, whether you cover a certain area. A chatbot that answers from your own approved information takes those off your phone and answers at 11pm when nobody is in. Make it clear to customers that they're talking to an AI assistant, not a person: since August 2026 the EU AI Act has required that. On a website we built for Kylemore Karting, their CFO Nicola Cosgrave put it simply: "The chatbot answers a lot of the questions people used to ring us for, any hour of the day."
Admin that follows the same pattern
Summarising meeting notes, pulling figures out of emails and invoices into a spreadsheet. It's repetitive work that follows the same pattern every time, and someone can check the result quickly. Once a job like this runs reliably, it can often be automated end to end: supplier invoices read from your inbox and entered into your accounts, for example, with a person only checking the ones that don't match.
Cleaning up messy data
Customer lists split across three spreadsheets and an inbox, the same contact entered twice, addresses typed five different ways. AI can help merge those lists, flag the duplicates and put everything into one consistent format. Because this is customer data, only use an AI tool your business has approved for personal information. Check what it produces before you rely on it, then use the clean version for everything else, including any AI you add later.
First drafts of everyday documents
Most of the writing in a small business follows a pattern you've done before. Give AI your notes and an old example, and it will produce a workable first version of:
- Proposals, quotes and tender responses, built from the last one that won
- Presentation decks for a pitch, a bank meeting or a team update, from your bullet points
- Replies to enquiries and responses to Google reviews
- Job ads, staff procedures and onboarding checklists
- Social media posts, newsletters and website updates
You still read it and fix it, because it doesn't know your customers, your prices or your tone the way you do. The saving is the blank page, not the judgement.
Where it doesn't help yet
Questions it hasn't been given the facts for
Every AI tool works from what it has been given, whether that's ChatGPT, Copilot or the chatbot on your website. Ask it for something it doesn't have, like a price, a delivery date, a refund rule or a figure from last year's accounts, and it can produce an answer that sounds right but isn't. In a draft you're going to check, that's a quick fix. In a quote, an email or a chatbot reply that goes straight to a customer, the customer takes it as coming from you. A website chatbot asked about a group discount you don't offer, for example, may invent one. So give AI your own facts to work from, check anything before it goes to a customer, and set any chatbot to pass questions it can't answer to a member of staff.
Hiring, performance and credit decisions
The EU AI Act classes AI used to recruit or select staff, monitor or rate their performance, or check an individual person's creditworthiness as high-risk, with extra legal obligations that apply from December 2027. That isn't limited to AI making the final call. The Act names AI that filters job applications or evaluates candidates, so a tool that sorts CVs or scores applicants is covered even if a person signs off at the end. For a small business, keep AI out of these decisions completely, including the filtering, scoring and ranking that lead up to them.
The habit behind the results
Trinity College Dublin and Microsoft Ireland's AI Economy in Ireland 2026 report shows how far some smaller firms are behind: they say only 25% of SMEs find AI saves two or more hours a week per employee, against 54% of large organisations.
The same report found one habit that made a clear difference. Organisations with a formal AI policy were ten times more likely to report major productivity gains, 30% against 3%. Only 44% had one, and OpenAI's research found 37% of Irish SMEs have no AI policy at all.
For a business of 15 people, the policy can fit on one page: which AI tools staff can use, which of them (if any) are approved for customer details or confidential information, and what AI is never allowed to tell a customer. Add a short briefing so everyone knows how to use the tools and where they go wrong. That also covers a duty the EU AI Act has placed on businesses using AI since February 2025: taking steps so staff understand the AI they work with.
Free template: a one-page AI policy
We've put that page together for you to download and fill in: approved tools, what never goes into them, telling customers when it's AI, where you never use AI, staff briefing and who to tell when something goes wrong. It's written around the EU AI Act's requirements for small businesses. It isn't legal advice.
Download the AI policy template (Word)How to start without wasting money
- 1Pick one job that eats hours every week and does little harm if the first attempt is wrong. Drafting proposals, writing up meeting notes or tidying a customer list are good places to start.
- 2Before buying anything new, check whether software you already pay for, like your email, office or accounting package, has an AI feature that does the job.
- 3Try it for a month and keep a simple tally of the time it saves. If the time isn't there, don't build it into how the business works.
- 4Once it's earning its keep, write down how it's used: which jobs, who checks the result, and what it must never be used for. That's the step that turns one person's shortcut into the way the business runs.
What getting it right gets you
- Hours back for the whole team, not just the owner
- Proposals, decks and replies that start from a draft, not a blank page
- Routine processes automated end to end, with a person checking the exceptions
- One clean customer list the whole business can trust
- Customers answered at any hour, from your facts, not AI guesses
- Staff who know which tools to use and what never goes into them
How we help
Choosing where AI fits, setting the rules for what it may and may not say, and checking it's still paying off three months later is part of our Virtual CIO service. If you want to see what that looked like on a real build, read the Kylemore Karting case study.
Not sure where to start? Book a free call and we'll pick the first task with you.
Sources: Opinium for OpenAI, survey of 200 Irish SME decision-makers (March 2026); Small Firms Association and Amárach Research, survey of 404 small firms (August 2026); Trinity College Dublin and Microsoft Ireland, AI Economy in Ireland 2026 (April 2026); McKinsey, The State of AI 2026 (survey of 1,719 respondents, May to June 2026); CSO, "Understanding of AI and its use in Ireland" (August 2026); EU AI Act, Regulation (EU) 2024/1689, Articles 4 and 50 and Annex III, as amended by the AI Omnibus (in force 27 July 2026).