The question usually comes from an owner or office manager who spends hours each week moving information from one place to another. Emails into a spreadsheet. Orders into an invoice. Form submissions into a CRM. The honest answer is yes, most of that can be automated without hiring an IT team. What it takes is one well-chosen process, a tool built for non-developers, and one person inside the company who owns the result.
The obstacle is rarely the technology. In Bitkom's September 2026 survey, 57 percent of German companies with 20 or more employees already used AI, and 85 percent of those that did not named missing technical know-how as the main reason [1]. Destatis found the same pattern a year earlier: 72 percent of companies that had considered AI but not adopted it cited a lack of knowledge [2]. The gap is know-how, not budget or ambition.
I. Start with the process, not the tool
Automation fails most often because a company buys a tool first and looks for a use afterwards. The better order is the reverse. Write down the tasks that repeat every week, who does them, and how long they take. Most lists contain the same few items: copying data between systems, sending standard replies, creating invoices or quotes, chasing missing documents, and compiling a weekly report.
A good first candidate has three properties. It happens often, at least several times a week. It follows rules a new employee could learn in an afternoon. And a mistake is cheap to fix. Lead intake from a website form meets all three. Payroll does not. Start where an error costs minutes, not money.
II. Map it before you build it
Before anything is automated, the process should fit on one page: what triggers it, which systems it touches, which decisions are made along the way, and what the finished result looks like. If two employees describe the same process differently, that difference has to be resolved first. Automating an unclear process only makes the confusion run faster.
This step also shows where the data lives. A process that depends on information in someone's head, or on a PDF that arrives in a different format every time, needs a small fix upstream before it can be automated. That fix is often worth more than the automation itself.
III. Choose a tool built for non-developers
Three categories of tools cover most needs. Hosted no-code platforms such as Zapier and Make connect common business apps through a visual editor and need no servers. n8n works the same way but can also run on a company's own server or in a European cloud, which keeps data inside the company and makes costs predictable at higher volumes. The third category is the automation already built into software the company pays for. Microsoft 365, Google Workspace, accounting and CRM systems often include rules and workflows that nobody has switched on.
For a company without developers, the deciding questions are practical. Does the tool connect to the systems already in use? Can the person who owns the process read and change the workflow? Where is the data processed, and is there a data processing agreement? A tool that passes those three questions is better than a more powerful one nobody in the company understands.
IV. Where AI helps, and where plain rules are better
Not every automation needs AI. When an order arrives, create an invoice: that is a rule, and a rule is cheaper, faster and more predictable than a language model. AI earns its place where the input is unstructured. Sorting incoming emails by topic, extracting fields from documents that never look the same twice, drafting a first reply, or summarising a long thread are all good fits.
The safest pattern for a small company is to let AI prepare and a person decide. The workflow reads the email, classifies it, drafts a response and places it in a queue. An employee checks it and sends it. Once the drafts are consistently right, the approval step can be reduced. Starting with full autonomy is how small errors reach customers.
V. What it actually costs
Software licences are rarely the expensive part. Among the companies already using AI in Bitkom's 2026 survey, 51 percent named infrastructure such as cloud and computing capacity as a major cost, 50 percent named preparing data, and 41 percent named integration into existing systems. Licences for AI software were a major cost for only 21 percent, and token usage for 8 percent [1]. Earlier in the year, a third of companies told Bitkom that AI had led to clearly higher costs than they expected [3].
The lesson for a company without an IT team is to budget for setup and upkeep, not only for subscriptions. A first workflow usually takes days rather than months to build. After that it needs someone to watch it: an alert when it fails, a monthly check that it still matches how the business works, and a plan for the day a connected tool changes its interface.
VI. The rules that cannot be skipped
Automation moves personal data between systems, so the GDPR applies from the first workflow. Every external tool that processes customer or employee data on the company's behalf needs a data processing agreement under Article 28, and the company should know where that processing happens. Running the workflow engine in the EU, or on the company's own server, settles many of those questions at once.
The EU AI Act matters less for typical back-office automation than its headlines suggest, but one point applies broadly. If an AI system talks to customers directly, such as a chatbot on the website, people must be told they are dealing with AI. Transparency duties for deployers applied from 2 August 2026, while the stricter high-risk obligations were pushed back under the 2026 Digital Omnibus agreement [4]. Legal uncertainty was the second most common reason German companies gave for not using AI [1]. It is worth one conversation with a data protection officer before launch. It is not a reason to wait.
VII. What automation does not solve
Automation makes a working process cheaper and faster. It does not fix a process that is broken, a team that disagrees about how work should be done, or data that is wrong at the source. It also does not remove the need for ownership. Every workflow needs one named person who notices when it stops working.
For a company of ten or fifty people, that is the real requirement. Not an IT department, but one process, one tool and one owner. Start there, measure the hours saved after a month, and choose the second process with that evidence in hand.
Batista builds AI and automation for European SMEs and startups, and audits the software they already run. Get in touch.
Sources
[1] Bitkom, "Erstmals nutzt die Mehrheit der Unternehmen KI," September 14, 2026.
[2] Destatis, "Nutzung von IKT in Unternehmen," survey results 2025, November 2025.
[4] Hogan Lovells, "EU legislators agree to delay for high-risk AI rules," via JD Supra, 2026.


