Telegram bots17 minJuly 10, 2026
Chatbot for support automation: FAQ, routing and working hours
Robert MirovUpdated Jul 10, 2026
"Are you open on Saturday?", "Where is my order?", "How do I pay with Payme?" — the same questions dozens of times a day. The manager copies answers from notes, the client waits in a message queue, and in the evening and on weekends some conversations simply go cold. A support chatbot is a messenger program (most often Telegram) that answers repetitive questions, accepts requests outside working hours, and hands hard cases to a human with full chat history. It is not a "replacement for the care team" — it is a filter and accelerator: the bot takes the routine; empathy and non-standard cases stay with the manager. Below is a working scheme for a business owner in Uzbekistan: what to cover with FAQ, how to build routing (escalation), what to tell clients at night, which metrics to watch, how to run a 1–2 week pilot, and what it usually costs in Tashkent. No promise that "the bot replaces half the staff".
01What a support bot is — and how it differs from a "sales bot"
Ads call everything a "chatbot", but the jobs differ. Separate two contours — or you will expect sales outcomes from a support tool.
A support bot answers questions from existing or almost-bought clients: order status, address, payment, returns, bookings. Goal — remove uncertainty quickly and keep trust.
A sales / lead bot qualifies "how much?", collects a phone number and writes to CRM. Goal — a lead. More on sales — sales automation; on the channel overall — why business needs a Telegram bot.
In practice, small business in Tashkent often uses one Telegram bot for both: FAQ first, then "leave a request" or "talk to a manager". The key is not to mix tone: in support the client is already annoyed by waiting or a failure; aggressive selling at that moment hurts.
| Parameter | Support bot | Sales bot | |
|---|---|---|---|
| Typical question | "Where is my order?", "How do I return?" | "How much?", "Any slot?" | |
| Success | Answer in seconds or escalation with context | Contact in CRM + qualification | |
| Risk | Menu loops, no path to a human | Pressure, spam after FAQ | |
| Metric | FAQ share without manager, reply time | Lead count, follow-up |
UZNEO usually builds a hybrid: FAQ buttons + honest escalation + lead capture when needed. AI (free text) comes after a knowledge base exists — see AI chatbot for website and Instagram.
02What the bot closes alone: FAQ and typical scenarios
FAQ (frequently asked questions) is not a pretty website section — it is a list of real client phrasing with approved answers. Most small-business support contacts are the same themes:
• opening hours, branch address, parking / landmark
• order, delivery or booking status
• payment methods (Payme, Click, Uzum, cash, transfer)
• delivery, pickup, return and exchange rules
• "how much is the base service / from what price"
• what to bring / how to prepare for a visit
The bot answers instantly with buttons or short text. The client does not wait while the manager finishes a call. Write answers in human language, not "press 1 for an operator". Build FAQ from chats of the last month (Telegram, Instagram Direct, WhatsApp), not from a meeting titled "what might people ask".
UZ example — food delivery / pharmacy: buttons "Order status", "Delivery zone", "Payme/Click", "Talk to a human". If the order is not found by number — escalate immediately, no ten-menu loop.
UZ example — clinic / salon: "Address and hours", "Price from…", "How to book", "Visit prep", "Manager". Diagnosis, treatment plan, individual discount — human only.
UZ example — online store: parcel status, ETA by district (Chilanzar, Yunusabad, Sergeli…), return within N days, size availability — if the system has no data, the bot must not guess.
If the question is not in the base — the bot says "passing to a manager" and does not loop the client. Golden rule: in 2–3 steps — an answer or a live human.
03Before and after: how the manager's day changes
Owners need to see routine change, not a feature list. Typical picture for a point with 20–50 messenger contacts per day:
| Before | After | |
|---|---|---|
| Manager pastes address and hours 30–40 times a day | Bot returns approved text in seconds | |
| Client writes at 22:00 — silence until morning | Question accepted; FAQ closed or ticket queued for 9:00 | |
| "Where is my order?" lives in one person's DMs | Status from the system or escalation to a group with history | |
| Complaint and "how much?" in one pile | Routing: price — bot; complaint — manager immediately | |
| Night requests lost at shift change | Logged in CRM or a ticket sheet |
This is not magic and not "+300% loyalty". It is less copy-paste and fewer lost night messages — if FAQ comes from real chats and escalation works.
04Routing: escalation table
Routing — rules for where the dialog goes: bot answers alone, calls a manager, or passes to sales. Escalation — handoff to a human. Support automation breaks when the bot cannot be left quickly.
Working flow:
1. Bot clarifies the topic: "Is this about an order, a payment, or something else?"
2. If the topic is complex or the client asks for a human — "Contact a manager" is always visible (not buried on level 5 of the menu).
3. Dialog lands in a support group with full history.
4. Manager replies in the same chat — the client does not repeat the order number and the issue.
| Topic | Bot | Manager | |
|---|---|---|---|
| Address, hours, "from" price | Answers | Updates the FAQ base | |
| Order / booking status | Answers if system data exists | Handles failures and "order not found" | |
| Refund, complaint, conflict, "I want my money back" | Escalate immediately | Owns the dialog | |
| Custom quote / B2B / large wholesale | Short qualify → human | Sales | |
| Medicine, legal question, "advise what to buy for a diagnosis" | Reference from base only, or human immediately | Specialist owns responsibility |
For a store with hundreds of orders per week this saves hours of routine per day (orientir). For a clinic or legal consult the bot filters "where are you" and "hours", and "need contract advice" goes straight to a specialist.
CRM integration (client and ticket tracking) logs the ticket even when the manager replies manually — so requests do not vanish in private chats during vacation or shift change.
05Working hours: what to tell clients at night and on holidays
The bot runs 24/7; the manager does not. Do not over-promise: clients in Tashkent often write after work, on Friday evening and on Sunday. The bot must accept and log the request even if a human replies in the morning.
Ready lines (adapt to your schedule):
• During hours (e.g. 9:00–20:00): "A manager will reply within 15–30 minutes. Urgent — tap Call."
• Off hours: "We received your question and will reply tomorrow from 9:00. Urgent — Call button / number on screen."
• Holidays / closed day: "No chat replies today; back on [date]. Emergency — call."
• Daytime peak: "There is a queue; the bot already logged your request, a manager will reply in order. Urgent — call."
Worst outcomes: silence, or endless "your request is important to us" with no action. Slightly better but still bad: promising "manager in 5 minutes" when there is no shift.
Night mode can take leads: simple question — FAQ; booking or order — intake with a morning manager alert. When support often flows into sales — see sales automation.
06How much load a bot really removes — and which metrics to watch
Without your stats — soft orientirs only, not a guarantee:
• at 20–50 contacts/day a bot often closes about half without a manager — if FAQ comes from real chats and phrasing matches how clients write;
• FAQ reply time drops from minutes (or hours at night) to seconds;
• one manager handles a larger flow when not copy-pasting the same lines;
• at 5–10 contacts/day payback is weaker: the bot still helps at night and on weekends, but "payroll savings" are almost invisible.
That does not mean cutting half the team. It means stop spending people on templates and keep them for conflicts, non-standard orders and upsells.
Weekly metrics (a simple sheet is enough):
• share of contacts closed by bot without escalation
• average time to first reply (including night)
• share of "talk to a human" right after a bot answer (if rising — weak base or irritating tone)
• number of tickets created off hours that got a morning reply
• factual errors (price, zone, ETA) — each one = FAQ fix
| Signal | What to do | |
|---|---|---|
| Almost everything escalates | Rebuild FAQ from the last 50 chats | |
| Clients tap "human" after the price answer | Answer too dry / incomplete — add 2–3 sentences | |
| Night tickets exist, nobody picks them up at 9:00 | Assign an owner and 15–30 min SLA from opening | |
| Complaints "bot didn't understand" | Add "human" on every step, simplify the menu |
If almost everything escalates — FAQ is poorly set, not "the bot is useless".
07Where it helps, where it irritates, and the launch checklist
Bot fits:
• delivery, food, retail, pharmacy — status, address, payment, zone
• service booking — slots (if calendar-linked), reminders, prep
• typical B2C with a predictable "from" price list
• online store with repeating "where's my tracking?" and "how do I return?"
Bot irritates:
• 12-button menus with no search and no "back"
• no fast path to a human
• "didn't understand" on a normal phrase
• fake live operator ("Alina is typing…") — clients feel it and get angrier
• outdated price and a promo that already ended
| Before (bad bot) | After (working) | |
|---|---|---|
| 8 screens to the address | Address in 1 tap from the main menu | |
| No "human" button | Escalation on every step | |
| "Didn't understand" on "open on Sunday?" | FAQ synonyms + hours button | |
| Promises a reply "now" at night | Honest off-hours template + ticket |
Launch checklist (1–2 weeks):
1. Export 30–50 real questions from chats (do not invent).
2. Build FAQ of 10–15 human answers; approve price and zones.
3. Escalation into a group with history; "manager" always available.
4. On/off-hours and holiday templates.
5. CRM or at least a ticket sheet (date, topic, status, who closed).
6. One pilot week: fix the base from 5 dialogs/day; log metrics before/after.
7. Assign an FAQ owner (one person) — or in a month the bot will lie.
Channel comparison: Telegram bot vs mobile app. Scenarios — UZNEO team and Telegram bot for business.
08Budget ranges and typical mistakes
Figures are a soft Tashkent orientir, not a UZNEO price list and not a package with a "fewer calls" guarantee. FAQ volume, order/warehouse integrations and CRM change the sum.
| Level | Typical scope | Orientir | |
|---|---|---|---|
| FAQ buttons | Menu, simple status, group escalation, hour templates | from ~1.5–3M UZS one-off | |
| Support + leads | FAQ + lead/ticket intake + CRM or sheet + pilot | ~3–6M UZS + support | |
| With system status | Link to orders / booking calendar | higher — by integration audit | |
| With AI support | Free text + strict knowledge base | higher; see AI bot |
A cheaper "bot in one evening" often fails on the first non-standard question: FAQ collection and pilot were not in the estimate.
Typical mistakes:
1. FAQ "from the head", not from chats — clients write differently, the bot misses.
2. No escalation — the bot argues with an angry client or loops the menu.
3. No ticket log — night requests vanish again.
4. AI on all channels without a base — hallucinations on price and zone.
5. Told clients "robot instead of support" — they expect miracles and get angry at humans.
6. Nobody updates price and zones — in a month the bot lies.
Budget not only button development, but base collection, pilot and one hour a week for fixes. Then the bot stays useful — not "another widget everyone bypasses".
Summary
A support chatbot is a filter and accelerator, not a full care-team replacement. It handles FAQ, accepts night requests without false promises, and passes hard cases to a manager with context — if you built the base from real chats, set escalation and watch simple metrics. Order: 10–15 real questions → bot/human table → hour templates → CRM or ticket sheet → one-week pilot → FAQ fixes → decide whether you need AI. Want less load on managers without quality loss — discuss with UZNEO or explore bot development.
Automate sales and leads?
Bots, CRM and integrations for your business. Automation plan in 24h on Telegram.
Related services
More articles
8
CRM22 min
Best CRM for business in Uzbekistan
CRM17 min
CRM for small business: what it is and why it matters in practice
CRM19 min
CRM implementation: stages, timeline and typical mistakes
CRM18 min
AmoCRM, Bitrix24 or custom CRM: how to choose
CRM19 min
CRM integration with Telegram, Instagram and your website: one funnel without lost leads
Telegram bots14 min
Why a business needs a Telegram bot: what it gives you and when to wait
Telegram bots19 min
Telegram bot vs mobile app: what a business in Uzbekistan should choose
Telegram bots18 min
