It's 11:47 PM. A prospect just finished dinner, picked up the tablet, and started looking for a divorce lawyer. They found you. Read the "about" page, checked the address, opened the contact form. They typed one line — "I'd like to book an initial consultation" — and hit send.
Your inbox will reply tomorrow morning at 9:15, when the assistant turns on the computer. Meanwhile the prospect has already opened three other tabs. One from a competitor who put a WhatsApp number on the site. Another from a firm with a chat on the homepage. By the third tab someone has already replied.
You're not losing that prospect at 9:15. You lost them at 11:48.
The 128-hour gap nobody counts
The average professional office in Italy is open about 40 hours a week: Monday-Friday, 9-13 and 14:30-18:30. When it closes, it stays closed for 128 hours — more than three times the open hours. Weekend included, your website is visitable 76% of the time no one is there to answer.
That gap wouldn't be a problem if every inquiry arrived during the "right" 40 hours. Reality is the opposite. A 2026 study of small-business calls measured that 62% of inbound calls go unanswered, and the highest concentration of missed calls is after 6 PM, on weekends, and in early morning hours. Applied to dental practices, the same research indicates that around 40% of patient inquiries arrive outside business hours.
For a lawyer, an accountant, a physiotherapist, a notary, the math is brutal. More than 40% of your prospects are trying to reach you when no one's there. If voicemail is the only answer, 85% of them never call back. 62% call someone else.
Why "they'll reply tomorrow" is no longer an answer
Twenty years ago, the patient who wanted to book a dental appointment waited. Not anymore. The Zendesk CX Trends 2026 report found that 74% of consumers now expect 24/7 service availability. 60% define "immediate" as a reply within 10 minutes. 90% say a fast response is "essential or very important" when they have a question.
This isn't a theoretical expectation. Sales-response-time studies show that replying within 5 minutes makes you 100 times more likely to convert an inquiry into a booked appointment. Not 2×. Not 10×. 100×.
The reason is psychological before it's commercial. When someone looks up a professional at 11 PM, they're not doing it out of curiosity. They just got a medical report, had a fight with their spouse, received a tax letter, decided they need to do something. In that 20-minute window they're ready to move. If you make them wait till morning, that urgency cools off. Often it disappears.
Traditional solutions and why they no longer cut it
Before talking about AI assistants, it's worth seeing why traditional tools don't cover this gap.
Voicemail hasn't moved in twenty years. It sends back a voice message that the caller listens to — if they listen — and asks them to call back during business hours. No confirmation, no answer, no booking. The abandonment rate is brutal: 85 out of 100 never call back.
The email contact form shifts the problem but doesn't solve it. The customer writes, gets an automatic "thank you, we'll be in touch" message, and waits. Meanwhile they've already opened other tabs. Average email response time in B2B professional services is 12 hours — well past the customer's patience threshold.
The outsourced answering service works better, but costs €200–600 a month for partial coverage and doesn't know your firm. The operator takes a name, a phone number, a short description. They can't tell the caller whether you handle that kind of case, what documents are needed, what your base rates are, when you have slots next week.
The rule-based bot — the kind everyone pitched in 2019 with button menus like "Press 1 for hours, 2 for contacts" — was a silent failure for many offices. The customer wanted to describe their problem, not click through a decision tree. The result was an awkward interaction that made the experience worse, not better.
What an AI assistant actually does (and doesn't do)
A modern conversational assistant, built on a language model like GPT or Claude, is something different. Not a rule-based bot. Not voicemail. It's a system that understands what the person is asking — even if they write it informally — and replies with information you've predefined.
On your website, at 11:47 PM, that prospect looking for an initial legal consultation might type in the chat: "what's the cost for an uncontested divorce and how do I book?". The assistant, if configured properly, replies with your initial-consultation fee range, explains that uncontested divorces typically require 3-4 sessions, and offers available slots next week. If the prospect confirms, it captures name and email and leaves you, the next morning, a near-closed lead.
Here's the key point, especially for regulated professionals: you decide what the AI can and cannot say. A lawyer can't give specific legal advice without an engagement. A psychologist doesn't diagnose by chat. An accountant doesn't sign off on tax positions in a message. A well-configured system has a system prompt that explicitly defines these limits: the assistant informs, orients, collects — but doesn't replace professional consultation. When the customer asks something that requires real advice, the AI says so openly and offers to book an appointment.
That's what separates a good assistant from a dangerous one. It's not the model's power — today's models are all very powerful. It's the frame you put around it.
The Italian context: numbers that make this urgent
According to the Politecnico di Milano's Artificial Intelligence Observatory, in 2025 the Italian AI market reached €1.8 billion, up 50% on 2024. But adoption is asymmetric: 61% of large companies have integrated structured AI solutions, against only 18% of SMEs.
That gap is the turning point. While your colleague across town still waits for "the right moment to go digital", their competitors are filling the 15-hour evening void where both of them are closed. The response gap is turning into a competitive gap.
Another figure worth attention: in Italy, WhatsApp is used monthly by 90.1% of internet users and is the preferred channel of 45.4%. 83% of Italian consumers say they're willing to reply to a WhatsApp message received from a company. Your client doesn't just expect to write anywhere, at any hour — they're already doing it with every other service provider in their life.
Three scenes at 3 AM
To see what really changes, it helps to imagine three concrete scenarios.
12:17 AM on a Thursday. A 42-year-old woman can't sleep. She's been thinking about a nose job for weeks. She opens the site of a plastic surgeon in her city, finds the chat, types: "how long is recovery after rhinoplasty?". The assistant replies that the most visible swelling lasts 10-14 days, that the surgeon operates out of a partner clinic, that the first consultation is €80. She asks if she can book the consultation for Friday the following week. She does. At 8:30 AM Friday, the assistant finds the request with name, contact, and chosen slot. The surgeon didn't do a thing.
Saturday afternoon, 4:20 PM. A man broke a tooth eating an almond. He's looking for an open dentist. Three practices don't answer. The fourth has an AI assistant that recognizes the emergency, explains that Monday's schedule is full but the practice holds urgent slots from 6:30 PM for 45 minutes, and offers to add the emergency to the agenda. Reassured, the man confirms. Monday at 6:30 PM he walks in with his broken tooth and a good impression of the practice. He won't call the other three again.
Sunday night, 9:35 PM. A business owner got a letter from the tax agency that he has to respond to within 60 days. He writes to an accountant he's heard about. The assistant explains that for tax notices the office follows a standard procedure, that the first assessment is free, that Monday mornings the accountant dedicates an hour to new inquiries, and offers a time. The owner confirms, feels heard, sleeps soundly. Monday the firm has a client already on the books — one who was about to write to two other firms too.
In all three cases, the professional didn't reply. The configuration did. And the client walking in the next day is a client who has already decided.
What to check before you turn one on
Not all AI assistants are equal. For a professional office, five criteria matter.
System prompt control has to be yours. If the provider doesn't let you edit what the AI says and doesn't say, you're buying a generic product. For a regulated professional, that's not a technical problem: it's an ethics problem.
Data residency has to be in Europe. Everything your clients type into chat is personal data. It has to be stored in EU infrastructure, with a DPA (Data Processing Agreement) that meets GDPR Article 28. The provider has to be able to supply one — otherwise you're importing the compliance problem.
AI Act compliance is a requirement from August 2, 2026. The user has to know they're talking to an AI. A good assistant already communicates this by default, with a clear opening message.
The ability to upload your documents is what separates a generic bot from a true office assistant. You have procedures, fees, specific areas of expertise. The AI has to be able to read your documents — bylaws, rate sheets, internal FAQs — and respond based on them.
Price has to stay sustainable even in a slow month. Today there are SaaS plans starting at €32/month for a complete setup. Less than your phone system, much less than an answering service, a fraction of what it would cost to hire someone just for after hours.
It doesn't replace anyone. It frees the right people for the right work.
A common fear is that an AI assistant will degrade the relationship with the client. The experience of the first practices that adopted them says the opposite. The assistant who doesn't have to reply 40 times a day to "what time do you open?" has more time to welcome the clients who walk in. The professional who finds three filtered, qualified requests at 9 AM starts the day with three appointments, not an inbox to empty.
AI doesn't replace the meeting. It replaces the void. The voicemail that said "we're closed, try tomorrow". The contact form that replied "we'll be in touch". The silence that, from the client's point of view, meant only one thing: there's no one here who's listening.
The difference isn't between a human office and an automated one. It's between an office that always answers and one that answers when it can. In 2026, the client picks the one that answers.
Want to see how an AI assistant would work for your practice, configured with your rules and procedures? Paste your website URL on iperchat.ai and try it in 30 seconds — free, no signup.