On average, businesses take 42 hours to reply to someone who has got in touch through their website. Almost one in four never replies at all. Harvard Business Review measured this by sending test enquiries to more than 2,200 companies, and the same study found something even more striking: businesses that reply within an hour are nearly seven times more likely to have a meaningful conversation with that customer.
That’s where the real value of an AI assistant lies, and it’s also the best way to work out whether the price makes sense. It isn’t about “having a chatbot”; it’s about making sure no interested customer is left waiting, whatever the time and whatever language they write in. A professional website assistant currently costs between €500 and €2,000 to set up, plus a monthly fee. Below you’ll find a full price breakdown and how to work out whether it pays for itself.
What an AI chatbot is today, and why it’s nothing like the old ones
We’ve all had to put up with an old-school chatbot: button menus that lead nowhere and answers that have nothing to do with the question. They ran on decision trees, so if your question hadn’t been anticipated, the bot didn’t know what to do. An AI assistant works differently. It’s built on a language model, the same kind of technology that powers ChatGPT, and it’s given your company’s knowledge so that it answers with your real information.
| Traditional chatbot | AI assistant | |
|---|---|---|
| How it understands the customer | Only if they pick an option from the menu | Understands questions however they’re worded, typos and all |
| Languages | One, or whichever are programmed in by hand | Replies to each person in their own language |
| Unexpected questions | Gets stuck | Finds the answer in your information or hands over to a person |
| What happens to the contact | Usually nothing | Saves it in the CRM with the conversation history |
| Follow-up | None | Can send reminders by email or WhatsApp |
The difference a business really notices isn’t the conversation itself, but what happens afterwards. A well-integrated assistant asks for contact details when it makes sense to, alerts the team when someone is interested and triggers an automatic follow-up if the person doesn’t take the next step. That’s the point where a chatbot stops simply “answering” and starts selling.
The channel matters too. In Spain, 92% of people aged 12 to 74 use WhatsApp, according to IAB Spain’s 2025 social media study, and 96% of them open it every day. For many customers, sending a WhatsApp message feels far more natural than filling in a form or waiting for a call back.
Where it really makes a difference to a business
In the assistants we’ve built, the impact shows up in three places: enquiries outside office hours, the team’s time and the leads that used to slip through the net for lack of follow-up. Two real cases show this well.
A tax advisory firm swamped by the same questions. A large part of the team’s day went on answering the same client queries over and over: deadlines, paperwork, how to complete a particular procedure. We set up an assistant with the firm’s information that handles those queries and only passes on what needs a professional. The result: 60% less time spent on support, with no drop in service quality. You can read the full tax advisory case study.
A language school with students from all over the world. Prospective students got in touch at all hours and in their own language, and the team couldn’t keep up. The AI assistant replies instantly in each student’s language and is connected to a CRM holding every contact’s record and history, plus 13 automated follow-up workflows.
As well as answering and following up, these are the uses that work best, based on what we see in our projects:
- Qualifying leads before they reach a salesperson. The assistant asks just enough to know whether the customer is a good fit and only passes on the ones worth pursuing.
- Appointments and bookings. It suggests slots, confirms them and sends reminders to cut down on no-shows.
- Order and delivery queries for online shops, which usually make up the bulk of enquiries.
- Picking up abandoned processes. It gets back in touch with people who started an application and never finished it.
What it should never do, and what the law requires
Knowing what an assistant shouldn’t do matters just as much as knowing what it can do, because that’s where your customers’ trust is won or lost. The number one risk is that it makes up answers: language models tend to say something even when they don’t know the answer. A well-configured assistant only answers with the information you’ve given it and, when it doesn’t have the answer, says so and offers to put the customer through to a person. Getting this right takes a solid knowledge base, precise instructions and plenty of testing with real conversations before it goes live.
Nor should it commit the business to anything. It can give guide prices or explain general terms, but it shouldn’t agree discounts, confirm lead times that depend on third parties or give advice that should come from a professional, such as a medical diagnosis or a legal opinion.
Mandatory from 2 August 2026Article 50 of the EU Artificial Intelligence Act requires people to be told they’re interacting with an AI system, unless it’s obvious. The regulation itself warns that a bot with a human name, a photo and a human tone shouldn’t be treated as obvious. In practice, a clear notice at the start of the conversation is enough, as long as the assistant isn’t dressed up as a member of staff. In our experience, customers don’t mind talking to an AI if it sorts out what they need; what they mind is feeling deceived.
And like any form on your website, the assistant has to comply with data protection rules, the difference being that a conversation usually collects more information. Before it goes live, check the following:
- A notice that it’s an AI at the start of the conversation.
- Privacy information that’s easy to find, and consent where required.
- Data stored on servers in the European Union wherever possible.
- Restricted access to conversations within the business.
- The option to speak to a person at any time.
How much does an AI chatbot cost for a business?
The price depends on what the assistant does and where it lives. These are the typical 2026 market ranges for set-up in Spain, compared with ours:
| Type of assistant | Market | MarketerosWeb |
|---|---|---|
| Basic website assistant (questions, lead capture, hand-off) | €500–€2,000 | €600 + €60/month |
| Assistant connected to a CRM and knowledge base | €2,000–€6,000 | Website + mini CRM: €1,500 + €110/month |
| Website and WhatsApp assistant with automations | €2,000–€8,000 | €1,900 + €120/month |
| Automated follow-up by email, or email and WhatsApp | Depends on the project | From €450 + €25/month |
All our prices exclude VAT, and you can combine them in our price calculator. Why is there a monthly fee? Because every conversation has a usage cost for the AI model, your business information changes (prices, services, opening hours) and the conversations need reviewing to spot the questions the assistant isn’t handling well. An assistant that isn’t maintained goes out of date within a few months, and an out-of-date assistant is worse than none at all.
How much would an assistant cost for your business?Choose website, WhatsApp, CRM and follow-up, and see the price instantly.
How to tell whether it’s worth it
Run two quick sums with your own figures. The first is about savings: how many hours a week your team spends answering the same questions. The second is about revenue: how many enquiries come in out of hours or don’t get a quick reply, and how many of them would turn into sales if someone answered straight away. Here’s an example with round numbers:
| Item | Example |
|---|---|
| Hours a week spent answering repeat questions | 5 hours |
| Team cost per hour | €20 |
| Monthly saving (if the AI handles 60%) | Around €260 |
| Customers won back each month by replying instantly | 2 |
| Value of each customer | €300 |
| Total monthly impact | Around €860 |
Set against a monthly fee of €60 to €120, the numbers clearly add up for most businesses with a reasonable volume of enquiries. Where it usually doesn’t pay off is in businesses that get very few enquiries, or where every enquiry is so different that it always needs a person.
Whenever we build an assistant, the process is always the same:
- Understand your enquiries. What people ask you, what information the assistant needs and what should happen to each contact.
- Set it up with your information and clear limits on what it can and can’t say.
- Test it with real conversations until it answers the way we want.
- Connect it to your website, WhatsApp, CRM and team alerts.
- Review and improve it every month based on the conversations.
If you’d like to see what an assistant could handle for you, take a look at our AI customer service and AI automation services, read the language school case study or tell us what kind of enquiries you get.
