It's eleven at night and a customer messages you on WhatsApp: "Do you have a slot tomorrow?". You don't see it until the next morning, and by the time you reply they've booked somewhere else. The next day, the same question about prices. And the day after, about opening hours. If that sounds familiar, it's not that you're bad at customer service: it's that you're answering the same thing a hundred times and messages still slip through. An AI assistant isn't magic, but it can help you stop losing those messages. Here's what it's really for and what it isn't.
Why customer service gets lost today
Hardly any business loses customers by serving them badly. It loses them by not replying in time. Enquiries don't respect your opening hours, and they arrive through three or four channels at once: WhatsApp, the website chat, Instagram, the phone. When one person has to cover it all, something falls through the cracks.
- Messages left unanswered outside hours. A customer asking at ten at night won't wait until morning: they message several businesses and go with the first one that replies.
- Repeated questions. Price, opening hours, availability, how to get there, how to book. The same dozen questions over and over, every single day.
- Slow replies at peak times. When work piles up, replying becomes the task that always gets pushed back.
- Nobody measures what's lost. Without a record of how many enquiries come in and how many get answered, the problem is invisible.
The outcome is the same in any sector: messages that go cold, customers who switch to a competitor and a team spending its time repeating information instead of dealing with what actually adds value.
What an AI assistant can (and can't) do
First, let's clear the ground. An AI customer service assistant isn't an employee, an oracle or something that "fixes everything on its own". It's a tool that answers based on the information you give it. Set up well, it can help you stay available when you're not and take the repetitive work off your plate.
- It can: answer frequent questions with your business's real information, stay available 24/7, handle several languages at once and collect a customer's details so you can follow up later.
- It can: filter enquiries and sort them by urgency or type, so your team gets to what matters without the noise.
- It shouldn't: improvise. If it doesn't know something, the right move is not to make it up: hand over to a person.
- It shouldn't: promise what your business can't deliver, or make sensitive decisions (complaints, legal or health matters) without human supervision.
The key is treating it as a support trained on your information, not a replacement. If you want to see the full range of what's possible, we cover it on our AI agents page.
Manual service vs. AI-assisted service
The quickest way to understand the difference is to compare them point by point:
| Aspect | Manual service | AI-assisted service |
|---|---|---|
| Opening hours | Limited to your opening hours | Also available outside hours |
| Repeated answers | A person repeats them every time | Resolves them on its own, from your info |
| Languages | Depends on your team's languages | Can handle several at once |
| Response time | Varies with workload and time of day | Instant for the questions it's prepared for |
| Escalation to a person | The team takes on everything | Passes sensitive or complex cases to a person |
The interesting part isn't choosing one or the other, but combining them: AI handles the predictable, and people handle what needs judgement.
How to set it up, step by step
You don't need a huge project. A useful assistant is built in phases, starting with the simplest part:
1. Define your frequent questions
Write down the ten or fifteen questions you repeat most. That's the base: without clear information, the assistant has nothing to answer from.
2. Connect WhatsApp and your website
Choose where the conversation starts. Starting with one channel — the one your customers use most — avoids complications.
3. Write the tone
Decide how it speaks: friendly or formal, in which languages and with what greeting. It should sound like your business, not a generic robot.
4. Turn on human supervision
Define when the conversation goes to a person: upset customers, complex cases or questions the assistant can't resolve.
5. Measure
Track how many enquiries come in, how many it resolves on its own and how many get passed on. That's what tells you what to adjust.
6. Refine
Real use reveals gaps: new questions, answers that could be better. An assistant gets sharpened over time.
Best practices and limits
A badly set-up assistant does more harm than none at all. These are the rules worth respecting:
- Don't promise what the business can't deliver. If the assistant says there's a slot on Sunday and there isn't, the problem isn't the AI: it's the information you fed it.
- Always escalate sensitive cases. Complaints, issues, health or legal matters should reach a person. Human judgement isn't negotiable there.
- Say it's an automated assistant. Customers have the right to know who they're talking to. Hiding it breeds distrust when you least want it.
- Take care of the data (GDPR). Collect only the information you need, and know exactly where it's stored and how it's used. Running an AI assistant means processing personal data, and that has to be done properly.
- Review what it says. Especially at the start. An odd tone or a wrong piece of data shows up when you read real conversations, not when you imagine cases.
In short: AI brings order and availability, but responsibility for what it says (and what it promises) stays with the business.
How to hire one
AI assistants aren't sold on a fixed rate, because no two businesses are alike. The same kind of assistant is built very differently if it handles a single question or covers bookings, quotes and support in three languages.
That's why at Solaris we build AI agents on a custom quote: you tell us your case, which questions you want to cover, through which channels and with what level of supervision, and we give you a fixed price before we start. No surprises on the invoice and no commitment until you're clear about it.
Do the maths: if the assistant stops you losing even a couple of customers a month by not replying in time, the time saved and the sales not lost can comfortably cover its cost. That's a direct return.
Why Solaris
We build AI assistants for real businesses, with no hype and no promises we can't keep:
- We start simple. A narrow case, one channel, and we grow from there. We'd rather deliver something that works than something that promises everything.
- Fixed, clear pricing. We tell you what it costs before we start and what's included, tailored to your case.
- We know the day-to-day of local businesses. We work with companies across Marbella and the Costa del Sol, and we know your time is limited.
- We tackle the limits head-on. GDPR, tone, human supervision: we discuss it from the start instead of leaving it in the air. You can see the rest of the AI solutions we offer.