Your agent's welcome message: the five lines that decide

Someone messages you at eleven on a Sunday night. They haven't seen your website, they don't know your team, and there's a good chance they're messaging two other local businesses at the same time. The first thing they read from you is your agent's welcome message — and those five lines decide whether they keep typing or close the app and go with whoever answers better. If you already have an agent replying on WhatsApp or your website chat, or you're about to set one up, here's what that welcome needs to say, in what order, and what it should never say.

What the customer is actually working out

They're not judging your prose. They're settling three questions in four seconds:

• Is anyone home, or is this a voicemail? If it smells like a canned auto-reply from ten years ago, their guard goes up and their expectations go down.
• Will they sort out my thing? Not "anything" — their thing, which is very specific and which they haven't described yet.
• So what do I type now?

Most welcome messages fail on the third one. They're warm, well written, deeply grateful for your contact — and they leave the person staring at a blinking cursor with nothing to say. And when someone doesn't know what to say, they don't ask. They leave. Writing the welcome is really writing the customer's opening line for them.

The five lines, one at a time

1. Who you are and what you are

The business name, and then, plainly, that they're talking to an assistant. Don't invent a "Laura" who doesn't exist — when the customer works it out halfway through, and they always do, you lose exactly the trust you'd just built.

It's also no longer purely a style choice. Since 2 August 2026 the transparency rules of the EU AI Act apply: if a system interacts directly with a person, it has to be clear they're dealing with an AI, unless that's obvious from the context. The European Commission's FAQ on Article 50 sets it out. One honest line handles the compliance and the trust problem at once.

2. What you can do — specifically

Three things, with nouns in them. "I can give you prices, check open slots and book you in." Not "I'm here to help with any enquiry," which means nothing and over-promises while it's at it.

That line does two jobs: it frames what's on the menu, and it teaches the customer your vocabulary. Say "quote" and people will write in asking for a quote, which your agent will parse first time.

3. The exit to a human

Say it before they have to ask: "if you'd rather talk to someone on the team, just say and I'll pass you over." It sounds counterproductive — aren't you inviting people to skip the agent? — but it works the other way round. Knowing the exit exists removes the fear of being stuck in a loop, and with that fear gone, almost nobody takes it straight away. The welcomes that hide the handover are the ones that get "I WANT TO SPEAK TO A HUMAN" in caps by message three.

4. One question, and a closed one

End on a question someone can answer in four words. "What do you need — a price, a booking, or something else?" Not "how can I help you today?", which drops them back at the blinking cursor.

One question, not three. Ask for the name, the car model and the registration in a single breath and you'll get one of the three back, then have to ask for the rest anyway.

5. When a real answer arrives

Setting the time expectation kills half the "hello??" messages that land twenty minutes later. "I reply straight away; if you need someone from the team, that's tomorrow from nine."

Put together, it reads something like this:

Hi — I'm the assistant at Rivas Motors (yes, an AI, saying so
up front).

I can give you rough prices, check slots to bring the car in,
and book you a time. If you'd rather speak to Jose or Marta,
say the word and I'll pass you over — they're in from 8:30.

What do you need: a price, a booking, or a question?

Five lines, eighty words, none of them decorative. It fits on a phone screen without scrolling, which is the only measurement that counts.

The hall of shame

Three welcome messages you see every week that actively cost you money:

The novel. Twelve lines, three emoji per line, company history and values thrown in. Nobody reads it — they scroll to the bottom and type what they were going to type anyway. All that copy did was slow them down.

The phone menu. "Reply 1 for quotes, 2 for bookings, 3 for problems." This drags the worst invention in telephony into a chat window, and when someone turns up with a thing that isn't 1, 2 or 3 — which is often — they fall out of the system entirely. If your agent understands plain language, let it.

The infinite promise. "Ask me anything, I'm here for whatever you need." What that means is "we never decided on the scope." Three messages later the customer asks for something the agent can't do, and the miss lands twice as hard because you promised it yourself.

Two more versions worth writing

Out of hours. The same message at eleven on a Sunday night should read differently. Not to apologise, but to set the expectation: "The workshop's closed, but I can get you a price and hold a slot — Jose confirms it first thing tomorrow." The worst thing you can do on a Sunday night is sound exactly like a Tuesday morning.

The customer who already knows you. If it's the fourth time this person has written, the full introduction is unnecessary and mildly insulting. "Hi again, Marta — is this about the Golf or something else?" Being recognised beats any introduction, and all it takes is the agent checking whether that number already exists in your customer record.

When to skip the welcome entirely

Here's the honest caveat, because there are three situations where a welcome message gets in the way.

The first and most common: when the customer opens with something specific. On WhatsApp that's nearly always. If someone starts with "hi, how much is a service on a 2019 Golf?" and your agent responds with five lines of introduction before answering, you've turned an answer into a form. The welcome is for people who arrive saying nothing, or just "hi." With a clear question in front of you, answer the question — you can introduce yourself along the way.

The second: low volume and a fast human. If four messages a day come in and you answer them yourself within ten minutes, putting an automated layer in front adds friction and saves you nothing.

The third: when the first contact is a delicate one. A complaint, a billing problem, an emergency. The agent can absolutely take the details and flag you, but a chirpy three-option welcome reads terribly in that moment. If it detects anger or urgency, short and straight to you is the right shape.

How to tell whether yours works

You don't need a dashboard. Pull thirty conversations from last week and count three things:

1. How many people replied to the welcome at all. If plenty read it and vanish without typing, the problem is line 4 — your question isn't inviting an answer.
2. How many answered with just "hi" or "hello". That's the tell that what's on offer wasn't clear. Fix it in line 2, with more concrete examples.
3. How many asked for a human within the first two messages. If it's a lot, either the agent is promising something it can't deliver, or the welcome sounds like a bot from a decade ago.

Rewrite based on what you find, then count again a fortnight later. It's one of the few automation jobs where half an hour of work shows up in next week's numbers.

Our AI receptionist ships with this structure by default and we adapt it with you — out-of-hours and returning-customer versions included. You approve how it sounds; it delivers that line a thousand times without getting bored.