For a hairdresser, an installer or a restaurant owner, the question about AI is not what it could do in theory. The question is which work it takes over, what stays with your team and what it costs to set up properly. In this overview we set out where AI pays off in a small business, where it disappoints and how to judge that for your own business.
What AI is good at and what it is not
Language models are strong at work involving language and fixed patterns: understanding a question, writing an answer from information you provide, summarising a conversation or pulling details out of a message. They are weak where the right information is missing. A model then picks what sounds plausible, and that is not always what is true.
Where it works
- Customer questions: opening hours, prices, availability and the status of an appointment or order, answered from your own data.
- Appointments and bookings: suggesting a free slot and putting it straight into the diary, in the evening and at weekends too.
- Follow-up: an appointment reminder, a request for a review or a message about a quote that is still open.
- Summarising and sorting: reducing a phone call or a long email to its essentials and passing it to the right colleague.
Where to be careful
- Amounts and commitments: discounts, guarantees and quotes belong with a member of staff or with fixed rules in your software.
- Advice with consequences: do not let a language model answer legal, financial or medical questions.
- Work without a source: if information is not recorded anywhere, the model will guess. Get your data in order first, then automate.
Customer contact is usually the first gain
In many small businesses most of the scattered time goes on the same questions by phone, WhatsApp and email. The hairdresser has scissors in hand when the phone rings, the installer is under a sink and the restaurant is in the middle of service. An AI assistant that answers at that moment and records an appointment or a callback request removes work exactly when it is most disruptive.
We have two add-ons for this: the AI assistant on WhatsApp, which answers questions, books and takes orders, and the AI phone assistant, which picks up when nobody can and sends you a short summary of each call.
Separate AI programs or AI inside your own software
Broadly, there are two routes. You can connect separate AI programs to your diary, till and mailbox, or you choose software that already includes the AI assistant.
Connecting separate programs
This is flexible, but every connection is a place where things can go wrong. The program only knows what you pass to it through that connection. If your price list or schedule changes, that information has to follow. And you manage several subscriptions, accounts and data processing agreements side by side.
AI as a module in industry software
We build software for each industry, including hair salons, hospitality, studios, hotels, trade businesses, car dealers and staffing agencies. The AI assistant is a module within it. It works with the same diary, the same customer records and the same prices as your team. A booked appointment therefore appears in the schedule straight away, and anything the assistant is not sure about goes to a member of staff.
What drives the cost
What AI costs depends less on the language model than on the work around it. These points make the difference:
- Size of the process: how many steps, exceptions and decision points there are.
- Integrations: how many systems it has to work with, such as your diary, till, accounts or mailbox.
- Quality of your data: how well prices, services and common questions are already recorded.
- Management after launch: use of the language model, checking the output and changes when your offer changes.
With our software you pay a one-off investment, which covers setting the software up fully for your business. After that you pay a fixed monthly licence for hosting, updates and support. You hear the price after the intake, because it depends on your industry and the modules you choose.
Measuring whether it pays off
You can only see whether AI pays off if you know where you started. So record the following for a few weeks:
- Time: how many hours a week your team spends on the questions or tasks you want to hand over.
- Response time: how long a customer waits for an answer on average, including outside opening hours.
- Missed contacts: unanswered calls and messages that were only picked up the next day.
- Corrections: how often a member of staff has to adjust an answer or booking afterwards.
Measure the same points again after you start. If time goes down but corrections go up, the assistant is not set up properly yet. How to run that first trial is explained in the step-by-step plan for your first AI automation. What AI can do in each department is covered in AI by business process.
Oversight is part of it
AI only pays off if you can trust it. That means deciding what the assistant may handle on its own and what a member of staff sees first, and reviewing the output regularly. How to set that up is described in human oversight in AI processes and checking AI output.
Frequently asked questions
Will an AI assistant replace my staff?
No. It takes over recurring questions and simple bookings, so your team is interrupted less often. Unusual questions, complaints and commitments stay with a member of staff.
Am I allowed to let AI process customer data?
Yes, as long as you handle it in line with the GDPR: a data processing agreement with the supplier, processing no more data than necessary and informing customers through your privacy notice.
How soon will I notice a difference?
That depends on how many recurring questions come in and how well your information is organised. If you measure beforehand, you will see it in your own figures after a few weeks.