--- title: "AI readiness for SMEs: a 10-question test on the foundations" description: "Before choosing which AI to use, check your data: ten questions to see whether your SME is ready for artificial intelligence. With Istat 2025 figures." --- [Home](https://se2marketing.it/en?hsLang=en) [What I fix](https://se2marketing.it/en#risolvo) [Connections](https://se2marketing.it/en#rete) [How I work](https://se2marketing.it/en#lavoro) [Pricing](https://se2marketing.it/en#offerta) [Blog](https://se2marketing.it/en/blog?hsLang=en) [About](https://se2marketing.it/en/about?hsLang=en) [Get in touch](https://se2marketing.it/en/contact?hsLang=en)[IT](https://se2marketing.it/blog) Blog # AI readiness for SMEs: a 10-question test on the foundations An SME is ready for artificial intelligence when its data is reliable, connected and kept up to date without manual work: before choosing a tool, it is worth checking that with ten questions. The starting figure makes the point: among Italian enterprises that ... **1 October 2026** 1684 parole · 9 min di lettura by **Sergio Selvelli** [← All articles](https://se2marketing.it/en/blog) **An SME is ready for artificial intelligence when its data is reliable, connected and kept up to date without manual work: before choosing a tool, it is worth checking that with ten questions. The starting figure makes the point: among Italian enterprises that considered AI without adopting it, 45.2% name the unavailability or poor quality of the data needed as an obstacle (Istat, the Italian national statistics institute, "Imprese e ICT", 2025 reference year, press release of 15 December 2025).** The question that comes up most often today is "which AI should we use?". It's a fair question, but nearly always a premature one. An AI model works on whatever it finds: if the same customer exists three times, if orders are retyped by hand, if nobody knows which system holds the right price, AI produces fast answers on wrong data. The test below is meant to show you where to start, not to give you a grade. Every "no" isn't a flaw: it's a lever. Fixing it makes the company faster even without AI, and when AI arrives, it works on data that holds up. ## What being "ready" for AI means By **artificial intelligence**, this article means systems that extract information from documents, generate text or answers, and classify or predict from company data. These are the most common applications among Italian enterprises that use AI: extracting knowledge from text documents (70.8%) and generative AI (59.1%), according to the same Istat survey. The **foundations** are the conditions that let those systems work on real data: a **single record** for each customer and product, a **system of record** for each type of data (the one that is right when two versions disagree), **automatic connections** between systems, and a **person responsible** for data quality. Without them, an AI project turns mostly into a data clean-up project, often one nobody budgeted for. ## How ready are Italian SMEs? Adoption is growing fast, but from a low base and with a clear gap between small and large companies. | Enterprises using at least one AI technology | 2024 | 2025 | | --- | --- | --- | | Italy, enterprises with at least 10 employees | 8.2% | 16.4% | | Italy, SMEs (10–249 employees) | 7.7% | 15.7% | | Italy, large enterprises (250+ employees) | 32.5% | 53.1% | | European Union, enterprises with at least 10 employees | 13.5% | 20.0% | *Sources: Istat, "Imprese e ICT", 2025 reference year, press release of 15 December 2025; Eurostat, news item of 11 December 2025. Universe: enterprises with at least 10 employees.* Among enterprises that considered AI without adopting it, these are the obstacles named (multiple answers allowed, so the shares don't add up): | Obstacle | Share | | --- | --- | | Lack of skills | 58.6% | | Lack of legal clarity | 47.3% | | Unavailability or poor quality of the data needed | 45.2% | | Privacy and data protection concerns | 43.2% | | High costs | 43.0% | *Source: Istat, "Imprese e ICT", 2025 reference year.* Data quality is the third obstacle, but it's the only one a company can solve entirely on its own, without waiting for the job market or the legislator. That's why the test starts there. The topic is covered in more depth in [AI in SMEs: the problem is the data](https://se2marketing.it/en/blog/ai-in-smes-the-problem-is-the-data?hsLang=en). ## The test: ten questions on the foundations Answer yes or no only. "Partly" counts as no. The third column shows where to work if the answer is no. | # | Question | If the answer is no | | --- | --- | --- | | 1 | Does each customer exist only once across your systems? | Single customer record | | 2 | Do you know which system has the final say on prices, stock and customer records? | A system of record for each type of data | | 3 | Do your CRM and ERP exchange data without anyone retyping it? | CRM-ERP integration | | 4 | Do online store orders reach the ERP on their own? | Online store–ERP sync | | 5 | For every order, do you know which campaign or channel it came from? | Tracking from campaign to order | | 6 | Does the dashboard you look at every month drive at least one decision? | A dashboard with few KPIs tied to decisions | | 7 | Is yesterday's data available today, without manual exports? | Automatic updates between systems | | 8 | Is someone responsible for data quality, even part-time? | Assign responsibility, don't buy software | | 9 | Are repetitive processes, such as orders, returns and quotes, written down somewhere? | Process mapping | | 10 | Do you know how many hours a month go into copying data from one system to another? | One week of measurement | ## How to read the result, and where to start A practical threshold, which is my own assessment and not a standard: **with fewer than six yeses, the first investment isn't an AI project, it's the foundations.** From six to eight, you can start with a narrowly scoped use case while closing the remaining "no"s. From nine up, the data holds and "which AI?" becomes the right question. 1. **Answer as two people.** Whoever runs the company and whoever works on the systems every day rarely give the same answers: the difference is already information. 2. **For each no, estimate the hours.** How many hours a month does that gap cost today? That number sets the priorities. 3. **Start with questions 1 and 2.** A single record and a system of record underpin almost everything else: connecting systems full of duplicates just duplicates the errors. 4. **Close one no at a time.** Each piece of work should produce a measurable result before moving on. 5. **Retake the test after three months.** The score is there to show progress, not to certify a level. Questions 1, 3, 4 and 5 already have practical guides: [a single customer record](https://se2marketing.it/en/blog/a-single-customer-record-how-to-build-one-in-an-sme?hsLang=en), [integrating CRM and ERP](https://se2marketing.it/en/blog/integrating-crm-and-erp-in-an-sme-where-to-start?hsLang=en), [syncing your online store and ERP](https://se2marketing.it/en/blog/syncing-your-online-store-and-your-erp-where-to-start?hsLang=en) and [tracking from campaign to order](https://se2marketing.it/en/blog/from-campaign-to-order-tracking-marketing-roi-in-an-sme?hsLang=en). ## When you don't need to wait for the foundations This too is my own assessment, not a data point: not all AI works on company data. A writing assistant for drafting emails or copy, or meeting transcription, can add value straight away, because it doesn't read customer records, orders or stock. The test is about projects that use the company's own data: forecasting, internal assistants, process automation. And a limit of the test itself: ten questions don't replace an analysis. They show where to start; they don't measure a company's maturity precisely. ## What it costs to fix the foundations There's no figure that fits everyone: it depends on how many "no"s there are and how many systems they involve. The structure of the spend is always the same. A **diagnosis** that checks the test answers against real data and estimates the hours lost; an **implementation** that closes the "no"s in order of priority; a **monthly review** that checks the data stays clean over time, because foundations degrade when nobody looks after them. ## Frequently asked questions ### Does the test work for a small company with few systems? Yes, and it's quicker. With only two systems some questions resolve fast, but questions 1, 2, 8 and 10 still apply to a ten-person company: duplicate customers, a system that's in charge, a person responsible and hours of retyping exist at every size. ### Why is the threshold six yeses out of ten? It's a practical threshold, not a statistical standard. Below six yeses, a single record and automatic connections are usually both missing, and an AI project would end up spending most of its budget on data clean-up. Read it as a signal of priority, not as a verdict. ### Can AI be used to clean up the data? Partly, yes: some tools help find duplicates or standardise addresses and codes. But deciding which system is in charge, who is responsible and which rules apply remains an organisational choice. AI speeds up the clean-up, it doesn't replace it, and a person still needs to check the results. ### How often should you retake the test? Every three months, or after any change to your systems: a new ERP, online store or CRM. The value of the test isn't the score at a given moment but the trend. If the yeses don't go up after a piece of work, that work didn't touch the foundations. ## In summary In 2025, 15.7% of Italian SMEs used at least one AI technology, against 53.1% of large enterprises (Istat), and 45.2% of those that considered AI without adopting it cite data as an obstacle. Before asking which AI to use, answer ten questions on the foundations: a single customer record, a system of record, automatic connections, tracking, responsibility for data. With fewer than six yeses, the first investment is the foundations. Every no you close makes the company faster straight away, and prepares the ground for AI that works on data that holds up. If you want to check the answers against your real data and see which "no" to start with, the diagnosis described on the [pricing](https://se2marketing.it/en?hsLang=en#offerta) page is built for exactly that. To talk it through: [get in touch](https://se2marketing.it/en/contact?hsLang=en). ## Sources Istat, "Imprese e ICT", 2025 reference year, press release of 15 December 2025 (universe: enterprises with at least 10 employees): use of at least one AI technology 16.4% in 2025 (8.2% in 2024), SMEs 15.7% (7.7%), large enterprises 53.1% (32.5%); most common applications among users: extracting knowledge from text documents 70.8%, generative AI 59.1%; obstacles among enterprises that considered AI without adopting it (multiple answers): lack of skills 58.6%, lack of legal clarity 47.3%, unavailability or poor quality of data 45.2%, privacy and data protection 43.2%, high costs 43.0%. Eurostat, news item of 11 December 2025: 20.0% of EU enterprises with at least 10 employees used AI technologies in 2025 (13.5% in 2024). The ten questions, the six-yes threshold and the guidance on reading it are the author's professional assessments, not research data. As far as could be verified, there is no official measure of Italian SMEs' AI readiness based on data quality. *Sergio Selvelli, 1 October 2026.* [← Previous**Syncing your online store and your ERP: where to start**](https://se2marketing.it/en/blog/syncing-your-online-store-and-your-erp-where-to-start?hsLang=en) About the author **[Sergio Selvelli](https://se2marketing.it/en/about?hsLang=en)** — I connect the CRM, online store, ad spend and data of small and mid-sized companies into one flow, and build automations and agents on top of it. Milan. [Get in touch](https://se2marketing.it/en/contact?hsLang=en) ## **Thirty minutes** to work out how to improve the systems you already have. [Get in touch](https://se2marketing.it/en/contact?hsLang=en)[Book 30 minutes](https://meetings-eu1.hubspot.com/sergio-selvelli) Marketing systems and operations for SMEs Milan Navigate [What I fix](https://se2marketing.it/en#risolvo) [How I work](https://se2marketing.it/en#lavoro) [Pricing](https://se2marketing.it/en#offerta) [Blog](https://se2marketing.it/en/blog?hsLang=en) [About](https://se2marketing.it/en/about?hsLang=en) Contact [Book 30 minutes](https://meetings-eu1.hubspot.com/sergio-selvelli) [Get in touch](https://se2marketing.it/en/contact?hsLang=en) [LinkedIn](https://www.linkedin.com/in/sergioselvelli/) [Instagram](https://www.instagram.com/se2marketing) SE² MARKETING · MILAN · REV.B 2026ONE FORM — SITE AND BLOG [Privacy Policy](https://www.iubenda.com/privacy-policy/25409837) [Cookie Policy](https://www.iubenda.com/privacy-policy/25409837/cookie-policy) ```json { "@context" : "https://schema.org", "@type" : "BlogPosting", "author" : { "@type" : "Person", "name" : "Sergio Selvelli", "url" : "https://se2marketing.it/en/blog/author/sergio-selvelli" }, "dateModified" : "2026-10-01T06:12:01.616Z", "datePublished" : "2026-10-01T06:12:01.000Z", "headline" : "AI readiness for SMEs: a 10-question test on the foundations", "mainEntityOfPage" : { "@id" : "https://se2marketing.it/en/blog/ai-readiness-for-smes-a-10-question-test-on-the-foundations", "@type" : "WebPage" }, "publisher" : { "@type" : "Organization", "logo" : { "@type" : "ImageObject", "url" : "https://se2marketing.it/hubfs/4.png" }, "name" : "SE² Marketing consulting" } } ``` ```json { "@context" : "https://schema.org", "@type" : "FAQPage", "inLanguage" : "en", "mainEntity" : [ { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "Yes, and it's quicker. With only two systems some questions resolve fast, but questions 1, 2, 8 and 10 still apply to a ten-person company: duplicate customers, a system that's in charge, a person responsible and hours of retyping exist at every size." }, "name" : "Does the test work for a small company with few systems?" }, { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "It's a practical threshold, not a statistical standard. Below six yeses, a single record and automatic connections are usually both missing, and an AI project would end up spending most of its budget on data clean-up. Read it as a signal of priority, not as a verdict." }, "name" : "Why is the threshold six yeses out of ten?" }, { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "Partly, yes: some tools help find duplicates or standardise addresses and codes. But deciding which system is in charge, who is responsible and which rules apply remains an organisational choice. AI speeds up the clean-up, it doesn't replace it, and a person still needs to check the results." }, "name" : "Can AI be used to clean up the data?" }, { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "Every three months, or after any change to your systems: a new ERP, online store or CRM. The value of the test isn't the score at a given moment but the trend. If the yeses don't go up after a piece of work, that work didn't touch the foundations." }, "name" : "How often should you retake the test?" } ] } ```