Use case

24/7 customer support without operators — setting up after-hours AI mode

Close 90% of overnight WhatsApp messages with AI, queue the complex ones for morning. Full setup: knowledge base, escalation rules, morning review.

April 19, 20264 min readBy: ZiFlow Team

Why this matters

A customer messages at 11:40 pm: "are you open tomorrow?". Next one at 6:15 am: "how much is service X?". Third at 2 am: "can I book for tomorrow morning?". Your staff are asleep. In the morning the owner sees 18 unread messages — a chunk of those people already booked with a competitor.

ZiFlow's after-hours mode answers 80–90% of these itself: hours, prices, address, available slots, order status. Complex cases are queued politely with "a team member will reach out first thing in the morning". The owner wakes up to 2–3 real cases, not 18.

90%
overnight messages closed without a human
< 15 sec
night-time response time
80%
of overnight customers retained

How night mode works

Three typical after-hours messages to a clinic:

  1. Simple fact — "what's your address?". Bot replies instantly: "22 Main St, 2nd floor. Open tomorrow from 9 am."
  2. Booking request — "can I come in tomorrow for a cleaning?". Bot checks Google Calendar, proposes slots, confirms. Customer wakes up to a confirmation.
  3. Complex case — "I have pain after the procedure, what do I do?". Bot: "That sounds serious. A clinician will call you first thing at 9 am. If it's urgent now, please call [emergency number]." A high-priority task is queued.

Step-by-step setup

Import your FAQ into the knowledge base

AI Agent → Knowledge Base → Upload. Markdown/TXT file or paste manually. Target 30+ Q&A pairs: prices, hours, address, services, refunds.

Configure escalation rules

AI Agent → Escalations. Instant handoff triggers:

  • Keywords: "complaint", "refund", "manager", "human"
  • Strongly negative sentiment (auto-detected)
  • Order value above $500
  • Same question asked a third time
Set operating hours

Settings → Operating Hours. Define team hours (e.g. Mon–Sat 9:00–18:00). Night mode activates outside them. Confirm the account timezone is correct (America/New_York, Europe/London, etc.).

Enable night mode

AI Agent → Night Mode → On. Opening template: "Hi! We're closed, but I'm [company]'s AI assistant — I can answer quick questions now and queue anything complex for the morning."

Auto-create morning review tasks

Settings → Automations → Add rule. "Night escalation → dashboard task at 9 am with priority overnight". Operators open "To handle" and blitz 3–5 items.

Set up an urgent channel

Settings → Notifications. Attach a Telegram channel or phone number. Add rule: keywords urgent, pain, emergency → SMS / call owner.

Morning review with corrections

Logs → Overnight conversations. For 3 weeks, spend 10 minutes each morning: flag wrong answers and click "Add correction to knowledge base".

Month-one report

Analytics → Night Mode. Metrics: % closed by AI, avg response time, optional overnight NPS. Target: 85–90% closure, under 30 sec response, NPS ≥ 8.

What good looks like

  • Closure rate: 80%+ of overnight conversations resolved without a human. Below 60% means the knowledge base is thin.
  • Response time: under 30 seconds. At night the customer doesn't wait — if there's no reply in a minute they bounce.
  • Morning task count: 2–5 per day for a typical SMB. More than 10 usually means escalations are too aggressive and simple questions are getting handed off unnecessarily.
  • Retained customers: compare "10 pm – 9 am" bookings before and after. Typical lift: +15–25%.

Before ZiFlow I'd wake up to 25 unread messages. Now it's 4, all real. The bot handles hours, address, pre-booking — I no longer burn the first hour of my day on chat.

David Lambert, dental clinic co-owner, Chicago

Common pitfalls

Keep reading

After-hours support compounds with auto-booking — overnight customers book slots instead of slipping away. Vertical playbook for salons: beauty industry guide. Side-by-side with a traditional chat widget that has no AI: ZiFlow vs Jivo.

Turn on after-hours AI mode — 14 days free
Try ZiFlow free
Share: Telegram WhatsApp

Use-case FAQ

What share of messages can the AI actually handle alone?
For a typical SMB (salon, clinic, shop) with a decent FAQ and clear pricing — 70–90%. Edge cases (complaints, refunds, custom orders) always escalate. ZiFlow doesn't try to fake expertise in hard cases.
Do customers know they're talking to a bot?
Your call. By default ZiFlow doesn't announce itself as a bot, but it doesn't pretend to be a specific human either — it answers on behalf of the company. If you prefer transparency, add 'I'm [company]'s AI assistant' to the first night-time response.
What if the AI gives a wrong answer?
Every night conversation is in **Logs → Conversations**. In the morning, the operator reviews, flags mistakes, and those cases are added to the knowledge base as corrections. Accuracy hits 95%+ within 2–3 weeks.
What about customers who ask 'connect me to a human'?
Instant escalation. The bot replies: 'Passing to a human agent, they'll respond by 9 am. For urgent matters call [number]'. A high-priority task is queued in the dashboard for morning.
Is there a load limit at night?
No. ZiFlow handles hundreds of parallel conversations — it's an AI, not a human. The only cap is WhatsApp's outbound rate-limit, and even 500+ nightly conversations fit comfortably under it.