The After-Hours Enquiry Map for Local Service Businesses

An after-hours enquiry workflow for a service business works best when every question is sorted into one of four lanes: answer now, collect and hand over, send to booking, or stop and escalate. The chatbot's job is not to sound clever. It is to recognise which lane a conversation belongs in and make the next human response easier.

A 9:42 p.m. enquiry

A customer visits a plumber's website at 9:42 p.m. with a leaking hot-water system. The business does not need an AI to pretend it has diagnosed the fault or to promise an arrival time. It needs a reliable first-response workflow: recognise the request, show approved safety wording, collect the right details and set an honest next step.

That distinction is the foundation of a useful after-hours assistant. It also explains why so many "24/7 chatbot" set-ups disappoint: they are configured to answer everything, rather than to route each kind of question to the right outcome.

If you have already read our guide to building a 24/7 enquiry workflow for missed calls, think of this map as the layer underneath it. It decides what the website assistant is allowed to do with each type of question before anyone writes a prompt.

The four lanes

Lane one: answer now

These are low-risk facts the business has already published: service areas, normal hours, services offered, accepted payment methods and the usual booking process. The assistant should answer only from approved sources. If a fact is not written down anywhere the business controls, it does not belong in this lane yet.

Lane two: collect and hand over

Quote requests belong here. Ask for a name, mobile number, suburb, job type, urgency, property type and a short description. For a visual issue, let the person say that photos are available rather than asking them to send sensitive or unnecessary material through chat. The goal is a record someone can act on first thing in the morning.

Lane three: direct to booking

Routine maintenance, consultations and non-urgent inspections can move to an approved booking link, but only when the business has deliberately made that path available. If the calendar is not kept current, leave this lane empty and use lane two instead.

Lane four: stop and escalate

Gas smells, electrical hazards, flooding near electricity, medical issues and anything involving immediate danger should never become an improvised AI diagnosis. Show the business's approved emergency wording and direct the visitor to the appropriate emergency service where applicable. This lane is short, but it is the one that protects the customer and the business.

Build the map in a spreadsheet

Create five columns:

Review the map with the person who actually answers the phone, not only the person who manages the website. Real wording from calls is more useful than a list of search keywords, because it shows how customers describe problems when they are stressed or in a hurry.

A few patterns usually appear once the map is filled in:

• Most after-hours questions are lane one or lane two. Hours, areas and "can someone call me back?" dominate.
• Lane four is small but non-negotiable. Write the exact wording once, get it approved, and do not let the assistant paraphrase it.
• Lane three depends on discipline. A booking link that offers times nobody can attend is worse than no booking link.

What a good after-hours reply looks like

For the hot-water leak, a reasonable reply might be:

"Thanks for letting us know. We're closed right now, so a person can't confirm a time tonight. If water is near any power points or switchboards, stay clear and call the appropriate emergency service. Otherwise, leave your suburb, mobile number and a short description, and the team will review it when we reopen."

It is honest about availability, applies the safety boundary, and collects exactly what the team needs. It does not guess the cause of the leak, quote a price or promise a time.

Implementation checklist

• Export 30 recent website enquiries, emails or call notes, and remove personal information first.
• Group them into the answer, capture, booking and escalate lanes.
• Write one approved response and one prohibited response for each high-risk topic.
• Configure the details you need to act on each enquiry.
• Test the after-hours wording on a phone, not just a desktop.
• Run ten awkward or adversarial questions before publishing.

Limits worth stating plainly

An enquiry map does not make a chatbot infallible. Generative assistants can still word things badly or miss context, which is why the high-risk lanes should rely on approved wording and a person. Accepting an enquiry at night is also not the same as attending at night: say which one your business offers. And if nobody reviews lane-two enquiries the next morning, the map has only moved the problem.

Building the map in Pivra

Pivra can be trained on your website, PDFs and other documents, or plain text you approve, and you can ask it questions in the dashboard to check its answers before it goes on your site. See how it works for the overall flow.

Each lane maps to a setting:

• Answer now: your training sources, with Strict mode on if answers must stay within approved material (see improving responses).
• Collect and hand over: lead capture and quote capture, so details land in your dashboard; see the quote request chatbot.
• Send to booking: a booking link to Calendly, a Google Calendar appointment schedule or another booking tool.
• Stop and escalate: approved wording as a refinement, plus human handoff for anything a person must decide.

Business hours let the assistant lead with your after-hours message outside opening times while still answering what it can. The after-hours lead capture page shows the end-to-end pattern.

A note on plans: the free plan (500 messages a month, one chatbot, no card) is a good way to test lane one, the answers. Lead capture, booking and human handoff are on Starter and above. Any paid plan can be tried free for seven days; a card is needed at checkout. Current details are on the pricing page.

Next step: fill in the four-lane map from last month's enquiries, then build the first version in Pivra and test every lane before it goes live.