15 July 2026

The virtual assistant that answers customers out of hours

A customer does not ask when your shutter is up. They ask when it occurs to them: at ten at night, on a Sunday morning, while waiting for the kettle before bed. If the first answer they get takes three days because it arrives on Monday at nine, there is a good chance they have found another option by then.

That is where an out-of-hours virtual assistant comes in: it replaces nobody on the team, but it stops a simple question from hanging all night.

What this kind of assistant does well

The cases where it works best are the repetitive, predictable ones: opening hours, indicative prices, how to get there, whether an appointment is needed, which documents to bring. Questions the team answers ten times a week with the same reply. A well-configured assistant can resolve them instantly, at any hour, without anyone having to look at their phone at midnight.

What a virtual assistant must never do

This is where you have to be strict, because mistakes here cost trust:

Never pass itself off as a person. The customer must know from the first message that they are talking to an automated assistant, not to someone on the team. Hiding it is not a minor detail: it is the kind of thing that, when discovered, does real damage to the relationship. Under the EU AI Act it is also, since August 2026, an obligation — I went through it in the piece on Article 50.

Never invent an answer. If the assistant does not know something — a price it has not been given, a medical or legal case outside its script — it must say so openly rather than improvise. An invented answer in a confident tone is worse than no answer.

Always have a clear way out to a person. Every conversation must be able to end with "the team will confirm tomorrow morning" or a direct contact number. The assistant is a first filter, not the end of the road.

An example: a dental clinic

A dental clinic was getting Instagram messages and missed calls out of hours, especially in the evenings, when people finally decided to call about the tooth that had been hurting for days. They put in a virtual assistant that answers basic questions (hours, whether they take emergencies, what a first visit involves) and that, faced with any specific clinical question ("it hurts, what can I do right now?"), does not venture medical advice: it takes a name and phone number and says someone from the clinic will call first thing. The result is not "selling more": it is that no enquiry goes unanswered for twelve hours, and the team arrives in the morning with the work already sorted rather than a chaos of messages to reply to.

Transparency and privacy are not optional

Two more things, beyond what it must not do: the customer has to be able to tell clearly that they are talking to an assistant (one line is enough; no heavy disclaimer needed), and the data collected — name, phone, reason for contact — has to be handled with the same care as any other health or personal data: stored securely, accessible only to those who need it, and never ending up training third-party models without anyone knowing.

It is not for everyone, and that is fine

An out-of-hours assistant makes sense when there is a real volume of repetitive questions outside opening hours. If a business gets three messages a month, it probably does not need one. The useful question is not "do I need a bot" but "how many questions go unanswered every week because they arrive when we are closed".

This article is general information, not legal advice for a specific case.