AI in a small business: the uses that actually work (and the ones that don't)
What AI can do for a small business today, without the hype: documents, meetings, customer service, data. And where it isn't worth it. From an IT & AI Manager's experience in manufacturing.
I work as IT & AI Manager at a manufacturing company. Every week someone asks me "can we do this with AI?". Sometimes the answer is yes and it saves hours. Other times it's no, or "yes, but it isn't worth it". This is my honest list for a small business.
What works today
1. Finding information in your own documents
Price lists, datasheets, procedures, contracts, old emails. In almost every company someone loses time every day looking for "where was that". An assistant that answers from your documents and says where each fact comes from saves that time.
The key is citing the source: if you can't check where an answer comes from, you can't trust it. I built an entire product on this, AlperMind, and today I apply what I learned in much simpler, cheaper versions.
2. Meetings that turn themselves into minutes and tasks
Record the meeting, transcribe it and extract the summary and tasks with owners. One of the clearest returns, because it removes a job nobody wants and that, when skipped, causes things to be forgotten. Details in meeting minutes with AI.
3. Classifying and summarising what comes in
Customer emails, incidents, orders, complaints. AI is very good at reading messy text and saying "this is an urgent complaint from customer X about order Y". That gets work routed faster.
4. Drafts, not final texts
Quotes, template replies, product descriptions, first drafts. It saves the blank page, but always with a person reviewing.
What isn't worth it (yet)
- Replacing a person in customer service. A chatbot that answers FAQs helps; one that tries to solve everything frustrates customers.
- Unsupervised decisions. Prices, supplier orders or anything involving money: AI can propose, a person decides.
- AI projects without a concrete problem. "We want AI" isn't a problem. "We lose two hours a day looking up prices" is.
- Building your own model. For a small business it almost never makes sense: existing models, used well, cover 95% of cases.
How to start without wasting money
- Pick a process that hurts: repetitive, text-heavy and done by hand today.
- Measure how long it takes now. Without the "before", you'll never know if it improved.
- Test small with tools you already pay for before building anything custom.
- Only if it works and falls short, consider a custom solution integrated with your systems.
Frequently asked questions
Is AI expensive for a small business?
For most uses, no. Many tools cost a few euros per user per month. What gets expensive is custom projects built before validating they solve a real problem.
What about the privacy of my data?
Choose the tool and its settings carefully: it shouldn't train on your data and, if data is sensitive, it should be processed in the EU or on your own servers.
Will AI replace my employees?
In a small business it's usually the opposite: it takes away repetitive work and frees time for what adds value.