What to do about inquiries that arrive after you close

Somebody rings your business at 7:10pm. Nobody answers, the call rolls to voicemail, and the voicemail is not checked until 9:40 the next morning, after the opening rush. By then the caller has rung two competitors and whichever one answered has the job. The same thing happens to the form submitted at 11:40pm on a Sunday and the text sent in the forty minutes between the last person locking up and anybody looking at the phone again. Nothing was handled badly. It simply was not handled at all for twelve hours, and the market does not wait twelve hours.

The 7:10pm inquiry is not a smaller inquiry

Most businesses answer inquiries from roughly nine to five and quietly lose everything else. The assumption underneath that is that out-of-hours contacts are casual browsers who will try again tomorrow. Usually the opposite is true. Someone contacting a business outside its own working hours is normally doing it after finishing their own working day, which means they have stopped browsing and started deciding. They are comparing two or three options in one sitting, and the deciding factor is frequently which one replies.

The failure is mechanical rather than attitudinal. Nobody chose to ignore the call. The routing simply had nowhere to send it after closing time, so it went to a mailbox that is a storage device rather than a response. The same is true of the web form that drops into an inbox nobody opens until mid-morning: the first human reply arrives half a day after the inquiry, long past the point where a reply still reads as service rather than an interruption.

Where the money actually leaks overnight

Three separate leaks run at the same time, and only one of them is obvious.

1. Response latency. Inquiries that get no acknowledgement for hours convert at a fraction of the rate of those acknowledged immediately. That gap is pure loss, because the lead was already paid for by whatever marketing produced it.
2. Attribution damage. When overnight inquiries never reach the pipeline, paid channels that run twenty-four hours a day look worse than they are. Budget then gets cut on the strength of a reporting artefact rather than on performance.
3. Quiet reputation attrition. A caller who reaches a dead voicemail twice stops calling. That decision is never recorded anywhere, so it never shows up as a problem to be solved.

There is a fourth, subtler cost: you have no idea how much demand arrives outside your hours, because the inquiries that were never captured were never counted. Decisions about staffing, opening times and media spend stay guesses until that volume becomes visible. Most owners are surprised by the number once they can see it, and a few discover their actual busiest inquiry hour sits outside the hours they open.

Why voicemail and a morning inbox check cannot fix this

The instinct is to work the backlog harder. Check voicemail first thing. Put someone on the inbox at 8am. Add an auto-reply saying you will be in touch. None of it closes the gap, for three reasons.

• A promise is not an answer. An auto-reply that says we have received your inquiry gives the person nothing to act on, so they keep shopping while they wait.
• A backlog competes with the morning rush. The overnight queue arrives at exactly the hour your team is least able to work it, which is why it routinely slips to mid-morning.
• Reconstruction is slow. Whoever picks up a voicemail has to call back, re-ask everything, and hope the person is available. The context the caller already gave is lost.

What closes the gap is not faster humans. It is a system that behaves, between closing and opening, as though somebody competent is awake: it acknowledges in seconds, has a short real conversation, and either books the appointment or hands a fully contextualised record to the morning team.

What good after-hours coverage looks like

1. An acknowledgement goes out within seconds of a missed call, a form submission or a chat message, from the same number the person actually dialled.
2. An AI assistant takes the thread and asks a small, fixed set of qualifying questions, one message at a time, in plain language.
3. It books into a real calendar with real availability, offering two specific times rather than a link, so the overnight inquiry arrives in the morning as an appointment rather than a task.
4. Urgency is detected before anything else happens, and genuine emergencies route to a named on-call person rather than waiting for the digest.
5. Anything the assistant cannot handle becomes a queued record with a written handover note, not a promise that someone is already on it.
6. Quiet hours are respected for anything non-transactional, with deferred messages queued for morning release rather than dropped.
7. At the start of the next working day the team opens a triaged digest instead of a voicemail box, with the urgent items first and the booked appointments already confirmed.

The full field list, pipeline design, escalation rules and workflow spec for this sit in the after-hours coverage build.

How to set up after-hours coverage, step by step

Step 1: Write down your business hours properly

Including the days you are genuinely closed and any seasonal variation. Every branch in this depends on it, and an inaccurate hours table is the single most common cause of overnight automation misfiring. If the hours table says you are open on a bank holiday, overnight inquiries that day get daytime treatment and nobody is there.

Step 2: Make the missed call text itself back

Use one number for both inbound calls and SMS, so the acknowledgement arrives from the number the person dialled. A text from an unfamiliar number reads as spam and gets ignored. Inside business hours the number forwards to your normal answering point; outside them, forwarding is bypassed and the missed-call route takes over instead of voicemail. If this is the only piece you implement, start with the missed-call text-back build.

Step 3: Stamp the timestamps that let you measure anything

Record when the inquiry arrived, whether it was outside hours, which channel it came from, and when the first outbound message went. Those four values are what turn response latency from an impression into a number. Without them you cannot tell a fifteen-second reply from a thirteen-hour one, and you certainly cannot show anyone the difference.

Step 4: Give the assistant four questions and a hard list of refusals

Decide in writing what it may ask, in order, and what it may never say. A sensible refusal list covers prices and fees, timelines, availability of specific staff, professional or clinical advice, and any commitment on someone else's behalf. It should also open honestly, naming that the office is closed and that it can take details and book, rather than pretending to be a member of staff.

Step 5: Name your urgency words and your on-call person

Write the urgency keyword list yourself. A generic list will miss the words your customers actually use and flag ones that do not matter in your trade. Then decide who is reachable overnight, by week, with a named fallback, and hold it on a thin on-call calendar so escalation always resolves to a real person rather than a role. Have the on-call person complete a short form afterwards, so the escalation leaves a record rather than a phone log.

Step 6: Open the first ninety minutes of the day wider

Overnight leads are the warmest leads you have, so let them take the earliest slots. Keep a separate set of short callback blocks too, so the assistant can book a time a human will ring rather than promising a vague morning call. A booked callback window is a commitment both sides can see.

Step 7: Build the morning digest before you launch

Group every overnight contact by what happened to it, with the ones needing a person listed first, each carrying a short handover note. Make the full conversation readable rather than summarized away, so whoever takes over does not re-ask questions the customer already answered at 9pm.

Mistakes that make after-hours automation worse than nothing

Mistake 1: Letting the assistant pretend to be a person

It gets found out, and the discovery costs more trust than the convenience gained. Naming the situation plainly tends to increase replies rather than reduce them, because people adjust their expectations and answer the questions.

Mistake 2: Stacking questions into one long message

Four questions in a single text reads as a form and gets ignored. One question per message, waiting for the reply, is slower on paper and faster in practice.

Mistake 3: Letting it quote prices

Pricing is the most common thing people ask about out of hours and the most expensive thing to get wrong. A quoted figure becomes a commitment in the customer's mind. Route pricing questions into a booked conversation instead.

Mistake 4: Sending marketing messages at 2am

Transactional confirmations are normally fine when someone has just booked. Nurture and promotional messages are not. Queue anything non-transactional for a morning release window rather than dropping it, and set quiet hours in the contact's time zone rather than your own. Rules differ by jurisdiction and industry, so confirm what applies to you with your own counsel before you launch.

Mistake 5: Inviting a reply you will not answer

If the overnight thread ends with an escalation promise and nobody picks it up by 9:30, you have converted a missed call into a broken promise. Agree who clears the queue, by when, and measure it.

Mistake 6: Letting the assistant collect things it should not

Overnight conversations gather personal information at a point when no human is supervising. Decide in advance what it may collect and configure it to refuse everything else, particularly payment details and identity documents, which do not belong in a text thread. Restrict who can read overnight transcripts, and review that access list when staff change.

How to measure after-hours coverage

Five numbers tell you whether this is working. Review them weekly for the first couple of months, then monthly.

• Median first-response latency for after-hours inquiries, from arrival to first outbound message. This is the headline figure and it should be measured in seconds.
• Overnight capture rate: contacts created out of hours divided by missed calls plus form submissions plus chat starts in the same window. A gap here means a channel is not wired in at all.
• Overnight reply rate, the share of acknowledged contacts who reply to anything. A low rate almost always means the first message sounds automated rather than that people are asleep.
• Booking rate without human involvement, and the clear rate by mid-morning for the records that did need a person.
• Show rate for overnight-booked appointments compared with daytime-booked ones. A meaningful gap means your confirmation ladder needs work, not that overnight leads are worse.

Add two more once the basics are stable: escalation response time, measured from the urgency match to the on-call task being completed, along with a count of escalations nobody picked up; and the volume of overnight inquiries by hour, which is the number that tells you whether your opening times match your demand.

Watch the opt-out rate on overnight threads as your brake. A rising number is the clearest available signal that the messaging has become pushy, and it is worth acting on before deliverability suffers. If you want this scoped against your own hours, channels and on-call arrangement, book a call and bring your hours table.

Frequently Asked Questions

Will customers mind talking to an AI assistant at 9pm?

Most do not, provided two things are true: it says what it is, and it is useful. People contacting a closed business already know nobody is at the desk. What annoys them is a dead end or a bot that loops. An assistant that acknowledges in seconds, asks sensible questions and offers two real appointment times is experienced as service. One that pretends to be staff, or that cannot answer anything, is not.

Should it be allowed to book appointments without anyone checking?

Yes, within boundaries you set. It should only ever offer slots from a live two-way synced calendar, so an external booking blocks the slot, and it should only offer services it is allowed to book. The risk is not autonomous booking, it is booking into availability that does not exist. Keep the harder cases, anything contractual and anything with an urgency flag on it, as a queued handover instead.

How long does after-hours coverage take to set up?

Typically one to two weeks of configuration, with two caveats on timing. Messaging registration for automated SMS can take longer than the build itself in some markets, so start that first and keep the SMS steps switched off until it clears. Unregistered traffic gets filtered silently, which is the worst possible failure mode for an overnight system, because the messages look sent and never arrive. And your hours table, urgency list, qualifying questions and on-call rota all need to be written down before anything is configured, because the system cannot invent them.

What happens to people who never reply?

They should not be deleted and should not be rung the next morning out of the blue. The sensible treatment is a light nurture sequence running in business hours, so the contact stays on the list and the inquiry stays recorded. They are counted in your overnight volume, which is part of how you find out whether your hours match your demand.