The goal is honesty, not omnipresence
Small teams lose more trust pretending to be 24/7 than admitting they are not. An "online" widget that answers at 9am tomorrow is a broken promise on every after-hours page view. The winning strategy is layered: be truthful about the humans, and make the non-human layers genuinely useful.
Layer 1: truthful state
Configure operating hours so the widget flips to its offline presentation the moment coverage ends. "Leave a message — we reply within one business day" converts surprisingly well, because it is a commitment rather than a bluff. The visitors you "lose" by showing offline were going to be lost at 9am anyway; the ones you keep now trust the button.
Layer 2: the self-serve night shift
Most after-hours questions are the same repetitive tier as daytime ones — hours, how-tos, policies. A public knowledge base surfaced inside the widget answers them at 2am with zero staffing, and an AI assistant trained on that same content handles the phrasing variations the search box misses. This layer is where after-hours actually gets resolved, not just captured.
Layer 3: capture with a promise
For everything the self-serve layer cannot absorb, the offline form is your contract: name, email, question, and a stated reply time. Two rules make it work: the promised time must be one you hit even on Mondays, and the auto-acknowledgment should restate the question so the visitor knows a human will see exactly what they wrote.
Morning triage: by wait, not by arrival
Overnight messages are already hours old — they are your oldest debt, and they compound. Start the day by clearing the offline queue BEFORE live chat opens, oldest first, and only then flip the widget online. Opening live chat first buries the overnight queue under fresh conversations and quietly breaks yesterday's promise to keep today's.
When to add real hours
Watch two signals in your transcripts and offline messages: after-hours volume that the AI could NOT absorb (judgment questions, purchase-blocking questions), and timezone clusters (a steady 3am block is a country, not insomnia). When either grows, extend coverage toward it deliberately — one evening block or one weekend morning — rather than defaulting to a heroic night shift nobody can sustain.
The MyLiveChat wiring
Operating hours, the offline form, the knowledge base in the widget, and the AI assistant with human handoff are all per-site settings on one flat plan — the night shift costs configuration, not headcount.
The overnight queue is a different queue
Messages left overnight are not simply daytime messages that arrived late. They have been waiting
for hours, some have already been solved by the customer, a few have escalated in the sender's mind,
and several have been sent twice through different channels. Treating them as a normal backlog is
what makes mornings feel worse than they need to.
Sort before you answer. The overnight pile usually splits cleanly into things that are now
irrelevant, things that are urgent and getting worse, and ordinary questions. Answering in arrival
order guarantees you spend your freshest hour on the first category.
Watch for duplicates across channels. The visitor who messaged at eleven and emailed at midnight
should get one coherent answer, not two partial ones from two people. Checking for a second contact
before replying costs seconds and prevents a specific kind of avoidable frustration.
The offline capture form is where most after-hours strategies quietly break, because it promises
in general terms and delivers in specific ones. “We will get back to you as soon as
possible” sets an expectation of hours in the reader's mind and often means tomorrow
afternoon.
Say when, concretely, in the visitor's terms rather than yours: the next working morning, within
one business day, on Monday if it is the weekend. If your coverage means a message left on Friday
evening waits until Monday, say that. It is better to lose the contact than to keep it on a promise
you will break.
Ask for only what you need to reply and to prioritise: a way to reach them, and enough of the
question to triage. Every additional field costs completions, and after hours you are already asking
for patience.
When after-hours is most of your week
The advice above assumes a staffed day with a quiet edge around it. For a large number of small teams that is backwards. If one or two people cover chat between other responsibilities, the unstaffed state is not the exception at the margins, it is the majority of the week.
That flips the design. When offline is the common case, the offline experience is your main experience and deserves the effort you would otherwise spend on the live one. The greeting a visitor reads at nine in the evening is seen by more people than the one your agents open with, and it should be written with the same care rather than left as whatever shipped.
It also changes what honesty requires. A widget that says nobody is available right now implies that somebody usually is. If that is untrue for most hours, the wording should describe the actual pattern instead: name the window you do cover, and say plainly what happens to a message sent outside it.
The failure mode to avoid is a chat presence that looks staffed and is not. A visitor who waits ten minutes for a reply that was never coming has been treated worse than one who was told immediately to leave a message. Running chat on a low-traffic site is mostly the discipline of getting that expectation right.
The shoulder hours behave differently
The hour before you open and the hour after you close are not the same as the middle of the night, and treating all unstaffed time identically wastes the most recoverable part of it.
A visitor arriving twenty minutes after close is usually still at their desk, still holding the question, and will very often still be reachable when you open. A visitor arriving at three in the morning is, from your side, a message in a queue and nothing more. The first is nearly a live conversation with a delay; the second is correspondence.
The practical move is to make the promise specific in the shoulder hours. Telling somebody at half past five that they will hear back first thing is a concrete commitment they can plan around, and it is one you can actually keep. The same sentence at three in the morning is vaguer than it sounds, because first thing may be eight hours of other people's messages away.
When you triage in the morning, the shoulder-hour messages are also the ones most likely to still be live opportunities. Working strictly oldest-first can mean answering last night's early-evening enquiry after the ones that arrived overnight, when the evening sender is the one most likely to still be waiting.
What to measure
Measure time to first reply on overnight messages from when the message arrived, not from when
your shift started. The second version flatters you and describes nothing the customer experienced.
Track how many overnight enquiries are still unanswered at the end of the first working hour. That
number, more than any average, determines whether your after-hours promise is real.
Watch the share of overnight contacts that turn into a second contact before you reply. A high
figure means your stated timeline is not being believed or not being met.
Whatever you decide about hours, check the evidence afterwards. The list of chats nobody answered is the most direct measure of an uncovered window, and it names the pages those visitors were on.