The After-Hours Lead Machine: AI Receptionist + Chatbot Setup That Never Misses a Call

Here's what happens between 5pm and 9am at most small businesses: a homeowner's pipe bursts, they call three plumbers, the first one who answers gets the $650 job. Your phone rang at 7:40pm. Nobody picked up. By morning the job is booked — with someone else. This playbook builds a system that never lets that happen again.

The anatomy of a missed lead

After-hours leads arrive in two forms: phone calls and website visitors. Calls go to voicemail (which 80% of callers won't leave). Website visitors hit your contact form and get silence until morning — by which time they've submitted the same form to two competitors. The fix is two answering layers working together: an AI receptionist for the phones, and an AI chatbot for the site, both feeding one inbox. Total budget: roughly $65–$190 a month. One saved job pays for the year.

Step 1: Choose your answering layer

Use this decision tree:

• Under 50 after-hours calls/month, already need a business phone: a bundled option like Quo Sona (~$15/user/month + ~$25 call packs) keeps it to one bill.
• 50–200 calls/month, high-value jobs: a standalone AI receptionist like Smith.ai (~$150/month for ~75 calls, ~$2 per additional call) with spam screening and human escalation.
• Callers are often stressed or negotiating (law, medical, high-ticket): budget for human backup — either a hybrid AI+human plan or a live service like Ruby, and accept the higher per-minute cost.

For the chatbot, you need a tool that answers from your content: Tidio's Lyro (~$29/month base + ~$39/month for 100 AI conversations) is the small-biz standard. Pick the receptionist first, the chatbot second, and make sure at least one of them can push leads into your CRM or inbox.

Step 2: Write the greeting script

The script is the whole game. A bad script makes the AI sound like a phone tree; a good one makes callers feel heard. Three templates to steal:

Trades (plumbing, HVAC, electrical): "Thanks for calling [Business] — this is [Name], the after-hours assistant. I can get you scheduled or get a technician the details. Is this urgent, or can it wait until morning?" Then: name, address, what's happening, best callback number, permission to text.

Professional services (law, accounting, consulting): "You've reached [Business] after hours. I'm [Name], the virtual assistant. I can take down the details of your matter and have [Owner] call you back first thing tomorrow. May I ask what this is regarding?" Then: matter type, timeline, contact details, conflict-check basics.

E-commerce / appointment services: "Hi, this is [Name] with [Business]. I can check availability, take your booking, or answer questions about orders and pricing. What can I do for you?" Then: intent (book, buy, support), details, contact.

Every script needs three non-negotiables: an AI disclosure line if your state requires it, a clear escalation path ("let me get a person on this"), and a hard stop on topics the AI must never improvise — pricing it hasn't been given, legal advice, medical guidance.

Step 3: Build the chatbot knowledge base from your FAQ page

A chatbot is only as good as what you feed it. Block out one afternoon and write answers to the 10 questions every service business gets: hours, service area, pricing/rates, how booking works, emergency policy, payment methods, warranties/guarantees, what's included in a visit, how quotes work, and cancellation policy. Each answer should be 2–4 sentences, end with a next step ("want me to book that?"), and include a handoff trigger — if the bot can't answer in two tries, it offers a human callback instead of guessing.

Feed it your actual FAQ page, your services pages, and your Google Business Profile Q&A. Then delete anything outdated before you connect it — confident wrong answers about pricing cost more than no answers.

Step 4: Route it all into one place

Two answering layers are useless if leads land in two silos nobody checks. Route everything — receptionist call summaries, chatbot transcripts, form fills — into one inbox or CRM. The minimum viable setup:

1. Receptionist → sends call summary + recording link to a dedicated email or Slack channel within 2 minutes of hangup.
2. Chatbot → logs transcript to your CRM (HubSpot, Pipedrive, even a shared Google Sheet) tagged "after-hours".
3. Automation (Zapier/Make, ~$20/month) → new after-hours lead creates a task: "call back by 9am," assigned to whoever opens.

Test the routing with a fake lead before you go live. The number-one failure mode of these systems isn't the AI — it's a summary email landing in spam.

Step 5: Test week — the 10 test calls to make before going live

Run these before any real caller hears the system:

1. A normal booking request. Pass bar: booked correctly, details accurate.
2. A pricing question. Pass bar: quotes only your real rates.
3. An urgent/emergency call. Pass bar: escalates, doesn't minimize.
4. A spam/robocall. Pass bar: filtered, not counted.
5. A caller who won't give their name. Pass bar: polite, still captures the issue.
6. A wrong number. Pass bar: identified and ended quickly.
7. Someone asking for the owner by name. Pass bar: takes a message, doesn't transfer blindly.
8. A caller speaking slowly or with an accent. Pass bar: asks to repeat rather than guessing.
9. Two questions in one call. Pass bar: handles both, doesn't drop the first.
10. The chatbot: ask the same pricing question on the website. Pass bar: same answer as the phone script.

Fix every failure before launch. A system that's 90% right still loses the 10% of callers it fumbles — and those are disproportionately the urgent, high-value ones.

Measuring it: the 3 numbers to watch in month one

• Answer rate: what share of after-hours calls got a real answer vs. voicemail. Target: 95%+.
• Capture rate: what share of answered calls produced a name + number + need. Target: 70%+. Below 50% means your script is leaking.
• Callback time: median time from AI handoff to human follow-up next morning. Target: under 60 minutes after opening. The AI buys you the lead; speed closes it.

Track these weekly for the first month, monthly after. If capture rate is high and closed jobs aren't moving, the problem is your follow-up, not the AI.