AI in your business: when it helps, when it’s a burden and how to start

At the last trade fair, three suppliers talked to you about artificial intelligence. Your bank invites you to a morning on AI. A competitor says on LinkedIn that they use it for everything. And you wonder whether you are falling behind, and where you would even start.

The pressure is real, but the useful question isn’t whether to use AI, but what for. AI is a tool, not a project: chosen well, it gives you hours back and improves your service; bought so as not to fall behind, it becomes a monthly expense nobody uses. This guide helps you tell one case from the other.

What the AI everyone talks about actually is

When people talk about AI today, they almost always mean software that understands and writes text, such as ChatGPT, Gemini or Copilot. It reads a document, summarises it, drafts a reply, sorts an email or answers a question in plain language.

It has two limits worth keeping in mind from the start. It knows nothing about your business beyond what you give it: without your data, it answers in general terms. And it isn’t always right: sometimes it makes up a fact or a figure with complete confidence, and the polished writing makes the mistake harder to spot.

Where it tends to help a mid-sized company

AI tends to help when it works with text and someone checks the result. These are the uses that pay off most in companies like yours:

• Finding what you already know. Asking your manuals, contracts, product sheets or procedures a question in plain language, instead of digging through folders or asking whoever has been there longest.
• First-line customer service. Answering the usual questions at any hour, such as lead times, order status or terms, and passing anything complicated to a person.
• Sales work. Drafting a proposal, summarising a meeting or tailoring a product sheet to each customer.
• Documents that arrive in a mess. Supplier invoices in different formats, orders written out in an email, scanned delivery notes.
• Asking your data. “Which customers have cut their orders this quarter?”, without waiting for someone to build a report. This needs your data to be tidy and connected.

If what you want is to remove repetitive work with clear rules, you often don’t need AI at all: ordinary automation is cheaper and more predictable.

When AI becomes a burden

AI becomes a burden when it’s bought before knowing what for. These are the most common signs:

• Licences nobody uses. A tool is rolled out to the whole company, and by the third month four people use it.
• More checking than saving. The results need so much checking that the task takes as long as before, or longer.
• Messy data. AI on incomplete or contradictory data gives bad answers, well written.
• The endless pilot. An impressive demo that never makes it into day-to-day work.
• A cost that grows by itself. Many tools charge per use, and without control the bill rises every month.
• Lock-in. Everything is tied to a closed tool, and if it changes its prices or terms, you have no alternative.

Your data and the law

With AI, the responsibility is still yours: for your data, for what goes out under your company’s name and for complying with the law.

• Your data. The free versions of many tools may use what you type to improve their models. Business versions usually let you switch that off and sign a data processing agreement, as data protection law requires. Don’t paste customer or employee data into a tool without that agreement.
• Your team already uses it. Someone in your company very likely already uses ChatGPT with work documents. Better to give them a safe tool and clear rules than to ban it and never find out.
• European law. The EU AI Act is already in force. Since 2 August 2026, anyone talking to an automated assistant must be told it is a machine. And if you use AI to make decisions about people, such as screening candidates or granting credit, the requirements are much stricter from December 2027.

Check the legal side with your adviser. The technical side, such as where the data lives, who can access it and what gets logged, has to be settled in the design, not bolted on at the end.

How to start without risking anything

You start with a small trial on one concrete problem, not with a tool for the whole company.

1. Pick a problem, not a tool. “We take two days to answer quotes” is a good starting point; “we want AI” isn’t.
2. Measure how it’s done today. How long it takes, how often it happens and how many mistakes there are.
3. Try it for a few weeks, in one area, with real data and with the people who will do the work. They know better than anyone whether it helps.
4. Decide with numbers. If it saves what you expected, roll it out. If not, stop. Stopping in time is a good result too.

What you can do this week:

• Ask your team who already uses AI and for what. It’s probably more people than you think.
• Find out which company documents are being pasted into free tools.
• Note three text tasks that repeat every week: emails, reports, proposals.
• Ask any supplier offering you AI for a real case from a company like yours, with what it saved.
• Decide who in your company will be responsible for AI.

How we approach it at vitamina.dev

Our job is telling you where, and where not. We start with your business, not the tool: we look at where time is lost or customers slip away, and we tell you frankly whether AI helps there, whether something simpler will do, or whether it isn’t worth it yet.

When it fits, we try it small, measure it and connect it to the software you already use, with your data protected and a person checking wherever there is risk. Our team has experience in AI integration and automation, management systems and financial platforms, where mistakes aren’t forgiven. If AI is going to be a burden for you, we’ll tell you before you spend anything.

Frequently asked questions

Does my company need artificial intelligence?

Not necessarily. It needs to solve its problems, and sometimes AI is the best way and sometimes it isn’t. If your company handles a lot of text, such as emails, documents, proposals or customer questions, it will probably help. If your problems are messy data or systems that don’t talk to each other, it’s worth fixing that first.

Is it safe to use ChatGPT with my company’s data?

It depends on the version and the contract. Free versions may use what you type to improve the service, so they are not the place for customer or employee data. Business versions usually let you switch that off and include a data processing agreement. Review it before giving your team access.

How much does it cost to bring AI into a mid-sized company?

It depends on what you want to solve. The cost usually has three parts: the initial work of preparing data and connecting the AI to your software, what the tool charges per use or per user, and maintenance. That’s why it pays to start with a trial of a few weeks and decide with numbers before investing more.