In short: An AI chatbot on your website, in Messenger or in WhatsApp helps when it takes on repeat questions, gathers the details an enquiry needs, books appointments and replies out of hours. It hurts when it hides the human, makes up answers or cuts short a conversation a salesperson should lead. Connect it to your enquiry process and CRM, measure results with a few clear numbers and, from 2 August 2026, tell people plainly that they are talking to AI.
Why are companies looking at AI chatbots right now?
Customers get in touch through several channels: the website form, Messenger, WhatsApp. Each channel is one more inbox someone has to watch. When a reply takes hours, some people may already have contacted another company.
The platforms are pushing in the same direction. At its Conversations event on 3 June 2026, Meta introduced Meta Business Agent, an AI agent for customer conversations on WhatsApp and Messenger that Meta is also extending to Instagram. For larger businesses, Meta also introduced the Business Agent Platform, which lets them build and customise their own agents. According to Meta, more than 1 million businesses already use Business Agent on WhatsApp and Messenger. That is Meta’s own figure, not independent research, but it shows where customer conversations are heading.
What does this mean for your business? The question is no longer whether conversations will involve automation, but what exactly it should take on and where it should stop.
When does an AI chatbot help?
An AI assistant is useful where a conversation has a clear purpose and a repeating pattern:
- Repeat questions: opening hours, address, delivery times, what a service includes, how returns work. Answers come from text the company has approved.
- Qualifying the enquiry: the chatbot asks what the salesperson wants to know before the first call, such as the type of service, timing, location or rough volume.
- Booking appointments: for clinics, beauty salons, garages and consultations, where the chatbot can see real availability.
- Out of hours: people get an answer, or at least a clear note on when someone will get back to them, instead of silence until morning.
The good news is that you do not have to start with AI. The WhatsApp Business app, for example, offers automated greeting and away messages, quick replies to common questions and labels for organising chats. For many small companies these features are enough as a first step, before an AI agent is even needed.
ScaleLab tip: Before choosing a tool, go through your last 100 customer conversations. Mark which questions repeat and which need a person. That is the chatbot’s job list, and the rest is work for the team.
When does a chatbot hurt?
This is the difference between an assistant that brings in enquiries and one that loses them:
- It hides the human: the customer wants to talk to someone, and the chatbot sends them round in a loop of questions. That customer often does not write again.
- It gives wrong answers: AI models can sound confident even when they are wrong. If the chatbot promises a price, deadline or discount that does not exist, the customer will expect the company to honour it.
- It leads complex sales: for an expensive service, negotiation or a tailored quote, the customer wants an expert. The chatbot can collect the details, but it should not replace the conversation.
- It collects data for no reason: every question in the chat is one more reason to give up. Ask only for what you actually use.
- It writes poorly in the customer’s language: unnatural wording tells the customer at once that nobody has checked the set-up.
How do you connect the chatbot to your enquiry process and CRM?
The chatbot is only the entry point. The value comes from what happens after the conversation. Our recommendation is to organise the process in these steps.
1. Define what counts as an enquiry
Decide when a conversation becomes an enquiry: a phone number or email left, an appointment booked, a quote requested. Without this definition you cannot compare chat with the website form.
2. Prepare the answer base
The chatbot answers from text the company has approved: services, terms, frequently asked questions. Do not let it promise prices, deadlines or discounts that are not in that text.
3. Set up handover to a person
Describe when the chatbot stops and passes the conversation on: when asked directly, for a complaint, for a question outside the answer base, for a large order. The person taking over must see the whole conversation so they do not ask the customer the same things again.
4. Record every enquiry in the CRM
Details from the conversation go automatically into your customer system, along with the channel (website, Messenger, WhatsApp) and the answers to the qualifying questions. The salesperson starts with context, and you can see which channel brings real customers.
5. Add the next steps
An email confirmation, an appointment reminder, a task for the salesperson if nobody has replied within a set time. These are the same rules we describe in building marketing automation.
What should you measure?
Without measurement you cannot tell whether the chatbot helps or just looks modern. Track at least these numbers every month:
| Metric | What it shows | What to watch for |
|---|---|---|
| Share of conversations that become enquiries | Whether chat brings in work | Compare with the same share for the website form |
| Share of conversations handed to a person | How much the chatbot cannot handle | Very high suggests a weak answer base, very low may mean a hidden human |
| Time to first response | How quickly the customer gets a reaction | Measure in and out of hours separately |
| Time to a human reply after handover | Whether handover works in practice | A long wait undoes the benefit of a fast chatbot |
| Enquiries that became customers | The quality of chat enquiries | Take the number from the CRM, not the chat tool |
Read the conversations themselves too. Going through 20 conversations a month reveals mistakes no dashboard catches.
What should you check on personal data and the AI Act?
Personal data
A chatbot collects personal data, so the same rules apply as for your website form. Tell people what you collect and why, cover the chatbot in your privacy policy and do not ask for more than you need. Do not encourage customers to share sensitive information in chat, such as health data, unless your process is set up for it. Also check where the chat tool stores conversations and whether it uses them for other purposes.
The duty to say it is AI
The EU Artificial Intelligence Act (AI Act) introduces transparency duties under Article 50, which apply from 2 August 2026. The amendments signed on 8 July 2026 postpone some rules for high-risk systems, but not these duties. For chatbots, this means people must be informed that they are interacting with an AI system, unless that is obvious from the context.
In practice: state clearly at the start of the conversation that an AI assistant is replying, and how the person can reach a member of staff. Do not give the chatbot a name and photo that present it as a real person. If you use an off-the-shelf tool, check that the provider lets you show a message saying the customer is talking to AI. There is more on the obligations in our article on the AI Act and what it means for companies. This is not legal advice: if in doubt, consult a lawyer.
ScaleLab tip: Use the first message for three things at once: say that an AI assistant is replying, what it can help with and how to reach a person. That meets the requirement and reduces frustration.
What should you do now?
Start by reviewing your conversations and listing what the chatbot takes on and what it hands over. Then define what an enquiry is, connect chat to your CRM and track the numbers from the first month. This is how we organise work on email marketing and automation: enquiry data is gathered in one place, and the repetitive steps that follow are handled by a system.
Frequently asked questions
Does an AI chatbot have to say it is AI?
Yes. From 2 August 2026 the AI Act requires people to be informed that they are interacting with an AI system, unless that is obvious from the context. We recommend stating it clearly in the first message, together with how to reach a person.
Can a chatbot replace the person who handles enquiries?
Not entirely. It can take on repeat questions, data collection and appointment booking, especially out of hours. Complex sales, complaints and tailored quotes stay with people.
Do you need AI for automatic replies in WhatsApp?
Not always. The WhatsApp Business app has automated greeting and away messages, quick replies and labels. For a small company with repeat questions, that is often enough as a first step.
How do I know whether the chatbot is working?
Track the share of conversations that become enquiries, the share handed to a person, response times and how many enquiries become customers. Compare with the website form and take customer numbers from the CRM.
Sources
- Meta: Introducing Meta Business Agent (3 June 2026)
- WhatsApp Business: Conversations 2026, Introducing Meta Business Agent on WhatsApp (3 June 2026)
- WhatsApp Business: Why small businesses should use WhatsApp Business app
- Gibson Dunn: EU AI Act Omnibus agreement
- Usercentrics: EU AI Act high-risk delay and Article 50 transparency