For the past two years, everyone has been talking to me about AI. A craftsman who wants ‘something with ChatGPT,’ a small business manager afraid of missing the boat, a boss who tried a chatbot once and didn’t understand how it could help him. The word is everywhere, yet in most of the small businesses I encounter, AI still hasn’t made any real impact.
I’ll try to be honest because this is a topic where empty talk flows freely. AI isn’t magic—it won’t ‘run your business’ for you. But when properly connected to your data and your tasks, it can take a significant portion of repetitive work off your shoulders. That’s what it means, in concrete terms.
General AI doesn’t know your business
When you use ChatGPT as-is, you’re talking to a model that has read a large portion of the internet but knows nothing about you: not your quotes, your product sheets, your customer interactions, or your internal procedures. It answers well in general, but misses the mark on your specific case. That’s why many executives try it, get disappointed, and conclude, ‘It’s not for us.’
The real leverage for a small business isn’t general AI. It’s AI connected to your own data.
Answering from your documents (what’s called RAG)
Imagine an assistant who has read all your documents: your technical sheets, your sales terms, your past quotes, your meeting notes, your internal FAQ. When a customer or employee asks a question, it answers based on those documents, citing the source—not on generic information found online.
This is called RAG (Retrieval-Augmented Generation). In short: the AI first finds the right information in your documents, then writes the answer based on that. The result isn’t a model making things up—it’s a model responding with your content.
For a small business, this makes a very practical difference: a new employee finding the answer on their own instead of bothering a veteran, a customer getting accurate information at 10 p.m., a pre-filled quote based on similar past quotes.
Doing tasks, not just answering (agents)
Answering a question is good. But the next step is doing the work. An agent is an AI you assign a multi-step task to, with permission to use your tools: reading an incoming email, extracting the info, filling out a form, drafting a response, and saving it as a draft for your approval.
The example I see working most often: sorting and preparing. Emails come in, the agent categorizes them, extracts attachments, fills in what needs to be filled in your tool, and presents everything ready for review. You keep the decision-making, and lose the chore.
The key word here is ‘draft.’ An AI well integrated into a small business shouldn’t act behind your back. It prepares, you validate. That’s what makes it usable without stress.
A concrete example
Take a 15-person SME that receives dozens of daily email requests: product questions, quote requests, customer service. Right now, someone manually sorts and distributes all of this.
With an assistant connected to their data: product questions receive a proposed answer, pulled from technical sheets, ready to send with one click. Quote requests are pre-filled based on similar past quotes. Customer service is prioritized by urgency. The person who used to sort doesn’t disappear—they shift from ‘copying and searching’ to ‘verifying and deciding,’ handling three times as many requests without losing their cool.
Custom-built or off-the-shelf tool?
Many small businesses don’t need custom development: an off-the-shelf tool with some built-in AI often suffices to start. I tell my clients this frankly. Custom solutions make sense when your data is sensitive (you don’t want to send it just anywhere), when your industry has rules standard tools can’t handle, or when you want the AI to integrate with your existing software—not the other way around.
In those cases, a solution built for you, hosted on your servers, connected to your tools, and following your rules, makes perfect sense. And it stays yours—not locked into a subscription you don’t control.
FAQ
Will my data go to an American tech giant?
Not necessarily. Depending on the sensitivity, you can choose models hosted in Europe, or even run certain parts on your own infrastructure. It’s a decision we make together, based on your constraints.
Do I need to be a large company for it to be worth it?
No. A small to medium-sized business with 10 to 50 employees, dealing with many repetitive questions or documents to sift through, is often the best fit: the time savings are immediate and obvious.
Will AI replace my employees?
Not with my approach. It handles the tedious tasks (copying, searching, sorting) and lets people focus on what they do better than a machine: deciding, reassuring, selling. AI prepares, humans validate.
How do I start without breaking the bank?
Start with one specific case—the one causing you the most trouble (sorting emails, handling customer service responses, searching through your documents). We’ll do it well, measure the results, and only expand if it proves its value.
Want to talk about it?
If you feel AI could help but don’t know where to begin, we can explore it together. I’ll start with an audit: which repetitive tasks cost you the most, what data you have, what’s realistic and what isn’t. I’ll provide a clear action plan and an honest budget. And if an existing tool already works, I’ll tell you.
