--- title: "Process automation in an SME: where to start without IT" description: "How an SME without an IT department should start automating processes: which ones to pick first, the four steps, and when it isn't worth it. Istat 2025 data." image: https://se2marketing.it/hubfs/copertina-09-process-automation-EN.png --- [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 # Process automation in an SME: where to start without IT In an SME without an IT department, process automation starts with a sheet of paper, not with software: you pick a frequent, repetitive process, write it down step by step as it is done today, time it for a week, and check that the data it starts from lives ... **15 October 2026** 1834 parole · 10 min di lettura by **Sergio Selvelli** [← All articles](https://se2marketing.it/en/blog) **In an SME without an IT department, process automation starts with a sheet of paper, not with software: you pick a frequent, repetitive process, write it down step by step as it is done today, time it for a week, and check that the data it starts from lives in one place only. Only then do you choose the tool. In Italy, 16.4% of enterprises with at least 10 employees use artificial intelligence, and among them about 18% use it to automate workflows (Istat, the Italian national statistics institute, "Imprese e ICT", 2025 reference year, press release of 15 December 2025).** Take order confirmation, a typical example. The order arrives by email. Someone keys it into the ERP, checks stock, prepares the confirmation and sends it to the customer. Every day, always the same way. Nobody has ever written it down: the person who has done it for years just knows how, and the exceptions (the customer on a special price list, the item that has to be ordered from the supplier) live only in their head. It is the ideal candidate for automation, and it is also where many automations get stuck. Software copies what it finds: if the process isn't written down, it copies a habit. This article opens the second level of the series, the AI ecosystem, and starts right here: the process first, then the tool. Done well, automation gives hours back to the people who currently spend them re-keying data. ## What automating a process actually means A **process** is a sequence of steps repeated for the same purpose: fulfilling an order, registering a new customer, chasing a payment. **Automating** it means handing to a system the steps that follow an "if X happens, do Y" rule, and leaving to people the steps that need judgement. The tools fall into three families: **rule-based automations** (workflows), which many CRMs, ERPs and online stores already include; **integration platforms**, which move data from one system to another; and **RPA** (robotic process automation), software that mimics a person's clicks on screen, useful when a system offers no other way to exchange data. AI comes in when a step requires reading free text, such as an order written in an email, and turning it into data. The distinction matters: most repetitive processes in an SME can be automated with simple rules, not with AI. AI is for the places where rules aren't enough. ## How many businesses use AI to automate? Official statistics only measure automation through artificial intelligence: rule-based automation has no dedicated indicator. | Indicator | Value | | --- | --- | | Enterprises with 10+ employees using at least one AI technology | 16.4% | | … among SMEs (10–249 employees) | 15.7% | | … among large enterprises (250+ employees) | 53.1% | | Among AI users: extracting information from text documents | 70.8% | | Among AI users: image recognition and workflow automation | about 18% (a single figure given by Istat) | | Among AI users: used to organise administrative processes | 25.7% | | Enterprises automating processes with rules, without AI | no official figure available | *Source: Istat, "Imprese e ICT", 2025 reference year, press release of 15 December 2025; population: enterprises with at least 10 employees. The "among AI users" shares refer only to enterprises using at least one AI technology. Istat reports image recognition and workflow automation as one approximate figure, and it is not made more precise here.* My reading, not Istat's: the most interesting number isn't 18%, it's 70.8%. The AI technology Italian businesses use most turns free text into data: exactly the step that, in many offices, is still done by someone re-keying. ## Which processes should you automate first? The test isn't how important a process is: it's how tedious it is. A frequent, repetitive process governed by clear rules is the first candidate, even if it looks minor. The table suggests some typical examples in an SME: they are examples, not client cases, and the prerequisite has to be checked case by case. | Process | Typical rule ("if X, then Y") | Prerequisite | | --- | --- | --- | | Order confirmation | If an order arrives and the goods are in stock, the confirmation goes out | Stock kept up to date in one system only | | New enquiry from the website | If an enquiry arrives, a contact is created in the CRM and a task assigned to sales | A rule against duplicate customer records | | Payment reminder | If an invoice is ten days overdue, the first reminder goes out | Correct due dates in the ERP | | Quote follow-up | If a quote has had no reply for seven days, a reminder is created | Quotes recorded in the CRM, not only in email | | Weekly report | Every Monday the week's numbers reach whoever decides | One source for every number | In every case the prerequisite is the same, seen from different angles: the data the automation starts from must live in one place, and it must be clear which system has the final say on it. That is the subject of the article on the [system of record](https://se2marketing.it/en/blog/system-of-record-in-an-sme-who-has-the-final-say-on-data?hsLang=en). ## Where to start without an IT department: 4 steps 1. **Pick the most tedious process, not the most important one.** Frequent, repetitive, with "if X happens, do Y" rules. Important processes are usually full of exceptions and judgement calls: they are the worst place to start. 2. **Write it down step by step as it is done today, exceptions included.** Not as it should be: as it actually happens. The person who does it writes it, with someone asking questions. List the exceptions one by one, because that is where an automation breaks. 3. **Time it for a week before changing anything.** How often it happens, how many minutes it takes, how many errors it produces. Without that number you will never be able to say whether the automation worked. 4. **Check that the starting data lives in one place.** If there are three versions of the same customer, automation multiplies them. In that case you start from a [single customer record](https://se2marketing.it/en/blog/a-single-customer-record-how-to-build-one-in-an-sme?hsLang=en), not from the tool. Only then do you choose the tool, and it often turns out you already have it: many CRMs and ERPs have automation features nobody ever switched on. If the process runs across two systems, such as the online store and the ERP, the work is the one described in [syncing your online store and your ERP](https://se2marketing.it/en/blog/syncing-your-online-store-and-your-erp-where-to-start?hsLang=en). Keep the sheet from step two: it becomes the automation's documentation. ## When automation isn't worth it This is my own assessment, not a data point. A process that happens three times a month shouldn't be automated: building, testing and maintaining the automation costs more than the time it saves, and the day it breaks nobody remembers how it worked. Nor is it worth automating a process that is still changing, because you lock in a rule that will last a few weeks. And it isn't worth it when the starting data is scattered across systems with no system of record: decide who is in charge first, then automate. In all these cases the right choice is a well-written manual procedure with a named owner. ## What automating a process costs There is no price that fits everyone: it depends on the process, how many systems it touches and how clean the data is. The structure of the spend, though, is the one described in [what integrating an SME's systems costs](https://se2marketing.it/en/blog/what-integrating-an-sme-s-systems-costs-the-three-parts?hsLang=en). **Analysis.** Write the process down, measure it, check where the data lives. It is the part many skip, and the part that decides whether the automation pays off. **Implementation.** Configure the tool, test the exceptions, go live. Sometimes an integration platform subscription is added. **Monthly review.** An automation nobody checks stops working silently: a renamed field, a software update, a rule that no longer applies. Someone needs to look at the error log. ## Frequently asked questions ### Do you need a developer to automate a process? To get started, rarely. Many CRMs, ERPs and ecommerce platforms include automation features you can configure without writing code, and the most common integration platforms work with building blocks. What you do need is someone who knows the process and has the time to write it down and test it. Custom development becomes necessary when systems can't exchange data automatically. ### Which process should you start with? The most frequent and most tedious one, with clear rules and few exceptions: order confirmations, registering enquiries from the website, payment reminders. Not the most important one. A simple first case that works teaches you more than an ambitious project stuck halfway, and gives you a real number of hours recovered. ### What's the difference between automation and AI? An automation carries out a rule decided by a person: if X happens, do Y. Artificial intelligence is needed when a step requires interpreting something, such as reading an order written in a free-form email. In most SMEs the first processes to automate can be handled with simple rules; AI comes later, where rules aren't enough. ### How do you know whether the automation worked? By comparing the numbers collected before you started, meaning repetitions, minutes and errors in the timed week, with the same numbers measured afterwards under the same conditions. Without the starting measurement there is nothing to compare, and the saving stays an impression. How to measure it properly is the subject of the next article in the series. ## In summary In an SME without an IT department, automation starts with a sheet of paper: the most tedious process, written down as it is done today, timed for a week, with its starting data in one place. Only then do you choose the tool, which you often already have. Start from the software and you automate habits. Start from the sheet and you automate the process. If you want to work out which process to automate first and how much time it gives back, the analysis described on the [pricing](https://se2marketing.it/en?hsLang=en#offerta) page starts from exactly that measurement. To check first whether the foundations hold, there is the [10-question test](https://se2marketing.it/en/blog/ai-readiness-for-smes-a-10-question-test-on-the-foundations?hsLang=en); on why AI depends on data, see [AI in SMEs: the problem is the data](https://se2marketing.it/en/blog/ai-in-smes-the-problem-is-the-data?hsLang=en). 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 (population: enterprises with at least 10 employees): at least one AI technology 16.4% of all enterprises, 15.7% of SMEs (10–249 employees), 53.1% of large enterprises; among enterprises using AI, extracting knowledge and information from text documents 70.8%, image recognition and workflow automation "about 18%", use in organising administrative processes 25.7%. As far as I could verify, there is no official statistic on how widespread rule-based automation is among Italian SMEs: this article says so instead of estimating it. The processes in the table are typical examples, not client cases; the selection criteria, the four steps and the "when it isn't worth it" section are the author's professional assessment, not research data. *Sergio Selvelli, 15 October 2026.* [← Previous**Systems that don't talk: a 7-point checklist for SMEs**](https://se2marketing.it/en/blog/systems-that-don-t-talk-a-7-point-checklist-for-smes?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" : "2026-10-15T06:02:41.277Z", "datePublished" : "2026-10-15T06:02:41.000Z", "headline" : "Process automation in an SME: where to start without IT", "image" : [ "https://se2marketing.it/hubfs/copertina-09-process-automation-EN.png" ], "mainEntityOfPage" : { "@id" : "https://se2marketing.it/en/blog/process-automation-in-an-sme-where-to-start-without-it", "@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" : "Article", "author" : { "@type" : "Person", "name" : "Sergio Selvelli", "url" : "https://se2marketing.it/en/about" }, "dateModified" : "2026-10-15", "datePublished" : "2026-10-15", "description" : "How an SME without an IT department should start automating processes: which ones to pick first, the four steps, and when it isn't worth it. Istat 2025 data.", "headline" : "Process automation in an SME: where to start without IT", "inLanguage" : "en", "mainEntityOfPage" : { "@id" : "https://se2marketing.it/en/blog/process-automation-in-an-sme-where-to-start-without-it", "@type" : "WebPage" }, "publisher" : { "@type" : "Organization", "name" : "SE2marketing", "url" : "https://se2marketing.it" } } ``` ```json { "@context" : "https://schema.org", "@type" : "FAQPage", "mainEntity" : [ { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "To get started, rarely. Many CRMs, ERPs and ecommerce platforms include automation features you can configure without writing code, and the most common integration platforms work with building blocks. What you do need is someone who knows the process and has the time to write it down and test it. Custom development becomes necessary when systems can't exchange data automatically." }, "name" : "Do you need a developer to automate a process?" }, { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "The most frequent and most tedious one, with clear rules and few exceptions: order confirmations, registering enquiries from the website, payment reminders. Not the most important one. A simple first case that works teaches you more than an ambitious project stuck halfway, and gives you a real number of hours recovered." }, "name" : "Which process should you start with?" }, { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "An automation carries out a rule decided by a person: if X happens, do Y. Artificial intelligence is needed when a step requires interpreting something, such as reading an order written in a free-form email. In most SMEs the first processes to automate can be handled with simple rules; AI comes later, where rules aren't enough." }, "name" : "What's the difference between automation and AI?" }, { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "By comparing the numbers collected before you started, meaning repetitions, minutes and errors in the timed week, with the same numbers measured afterwards under the same conditions. Without the starting measurement there is nothing to compare, and the saving stays an impression. How to measure it properly is the subject of the next article in the series." }, "name" : "How do you know whether the automation worked?" } ] } ``` ```json { "@context" : "https://schema.org", "@type" : "FAQPage", "inLanguage" : "en", "mainEntity" : [ { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "To get started, rarely. Many CRMs, ERPs and ecommerce platforms include automation features you can configure without writing code, and the most common integration platforms work with building blocks. What you do need is someone who knows the process and has the time to write it down and test it. Custom development becomes necessary when systems can't exchange data automatically." }, "name" : "Do you need a developer to automate a process?" }, { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "The most frequent and most tedious one, with clear rules and few exceptions: order confirmations, registering enquiries from the website, payment reminders. Not the most important one. A simple first case that works teaches you more than an ambitious project stuck halfway, and gives you a real number of hours recovered." }, "name" : "Which process should you start with?" }, { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "An automation carries out a rule decided by a person: if X happens, do Y. Artificial intelligence is needed when a step requires interpreting something, such as reading an order written in a free-form email. In most SMEs the first processes to automate can be handled with simple rules; AI comes later, where rules aren't enough." }, "name" : "What's the difference between automation and AI?" }, { "@type" : "Question", "acceptedAnswer" : { "@type" : "Answer", "text" : "By comparing the numbers collected before you started, meaning repetitions, minutes and errors in the timed week, with the same numbers measured afterwards under the same conditions. Without the starting measurement there is nothing to compare, and the saving stays an impression. How to measure it properly is the subject of the next article in the series." }, "name" : "How do you know whether the automation worked?" } ] } ```