To build a simple AI chatbot for a small business, decide the one job it should do, write a short knowledge base of facts it may use, pick a channel (website chat or WhatsApp), and set it up in a no-code tool such as Botpress, Zapier or n8n. Add a handoff to a person, test with real questions, then launch.
Why does a small business need a chatbot now?
Your customers are already messaging you. DataReportal counted 117 million internet users in Pakistan at the end of 2025, and 194 million cellular mobile connections active in late 2025 (Digital 2026: Pakistan). Many of them ask the same questions on WhatsApp and Instagram, often late at night when nobody is there to reply.
A slow reply loses the sale to the shop that answered first. A chatbot answers the routine questions in seconds, collects the details of anyone who wants to buy, and leaves your staff free for the conversations that need a person. It also keeps working on Eid holidays and at 2 a.m.
The cost of trying has also dropped. Chatbot builders have free plans, and on the WhatsApp Business Platform, replies inside the 24-hour window that opens when a customer messages you are free (WhatsApp Business Platform pricing).
What to decide before you build
Most chatbot projects fail on decisions made in a hurry, not on the technology. Answer these seven questions on paper first.
- Rule-based or AI? A button menu handles a few fixed choices. An AI chatbot understands free-text questions. The table below compares them.
- What is its one job? Pick the job that takes the most staff time today, such as answering questions or capturing leads.
- Which channel? Website chat, WhatsApp, or both. Go where your customers already message you.
- Who takes the handoffs, and when? Name the person and the hours. A handoff nobody answers is worse than no chatbot.
- What volume and budget? Estimate conversations per month, then check plan limits and AI costs against it.
- What data will it collect? Decide what you ask for, where it is stored, and how you tell customers.
- How will you know it works? Choose two or three numbers to track from day one (see the last section).
Rule-based or AI chatbot?
Rule-based chatbot | AI chatbot | |
|---|---|---|
How it works | Buttons and fixed replies you write | Reads free text and answers from your knowledge base |
Handles unexpected questions | No, it shows the menu again | Yes, within the limits you set |
Risk of wrong answers | Very low | Real; needs firm instructions and testing |
Setup effort | Low for a few options | Low to start, ongoing review |
Good for | Opening hours, menus, order tracking links | Varied questions in English, Urdu or Roman Urdu |
Many businesses mix the two: buttons for the common paths and AI for everything else. If you only get five kinds of questions, start rule-based and save the AI for later.
Step 1: Give the chatbot one clear job
The chatbots that fail are the ones asked to do everything. Start with a single job that takes up staff time today:
- Answering common questions: timings, location, prices, delivery areas, return policy.
- Capturing leads: collecting name, phone and what the person needs, then passing it to sales.
- Booking: offering available slots and confirming an appointment.
- Order status: looking up an order and giving its delivery status.
Answering common questions is the easiest first job, because it needs no connection to other systems. Order status and booking come next, once the basics work.
Step 2: Write the knowledge base
An AI chatbot is only as good as the facts you give it. Write a plain document with everything it is allowed to say. For a small shop or clinic:
- Business name, address, opening hours and contact number.
- Products or services with current prices, or a clear note that prices are shared on request.
- Delivery areas, delivery times and charges.
- Payment methods accepted, such as cash on delivery, JazzCash, Easypaisa or bank transfer.
- Return, exchange and refund rules.
- The 20 questions customers ask most, with the answer you would give.
Go through your WhatsApp and Instagram messages from the past month to find the real questions. Update this document whenever prices or policies change, or the chatbot will give out old information.
Step 3: Choose the channel
Website chat
A chat widget on your website is the simplest start. Most chatbot tools give you a snippet to paste into your site. It suits businesses whose customers find them through Google or ads.
In Pakistan, many customers would rather message on WhatsApp than use a website. A WhatsApp chatbot runs on the WhatsApp Business Platform, usually through a provider your tool connects to. Meta charges per message for some categories, but when a customer messages you first, a 24-hour customer service window opens in which service replies are free (WhatsApp Business Platform pricing). Start on your website if you want to learn cheaply, then add WhatsApp.
Step 4: Choose a no-code tool
- Botpress: a dedicated chatbot builder. Its free plan includes a small number of conversations a month on web chat, with WhatsApp on paid plans (Botpress pricing).
- n8n: its Chat Trigger node starts a chatbot workflow, using either n8n's hosted chat page or a chat widget embedded on your website (n8n Chat Trigger docs). It suits you if you want the chatbot to also update sheets, CRMs or order systems.
- Zapier: its free plan includes basic access to its AI chatbot and agent products, handy if you already automate with Zapier (Zapier pricing).
Plans and limits change often, so check current pricing against the number of conversations you expect before you build.
Step 5: Write the instructions
Every AI chatbot tool has a place for instructions, often called a system prompt. Keep them short and firm:
- Who it is: "You are the assistant for [business name], a [type of business] in [city]."
- What it does: "Answer customer questions using only the knowledge base."
- What it must not do: "Never invent prices, stock, delivery dates or discounts. Never promise anything not in the knowledge base."
- When to hand off: "If you do not know, or the customer is upset, or asks for a person, collect their name and number and say a team member will reply."
- Tone: "Friendly, short replies. Reply in the language the customer uses."
Step 6: Add a human handoff
A chatbot that traps an angry customer does more harm than no chatbot. Make sure every conversation can reach a person. The simplest version sends the conversation summary and customer number to the owner by email or WhatsApp, and tells the customer when to expect a reply. Then make sure someone actually replies.
Step 7: Test before launch
Ask a few friends to try to break it. Test with:
- Common questions, phrased in different ways.
- Questions the knowledge base does not cover, to check it hands off instead of guessing.
- Messages in Roman Urdu and mixed English, if your customers write that way.
- Requests for discounts or promises it should refuse.
- Rude or confusing messages.
Fix the knowledge base and instructions after each round, and test again until it behaves well.
Step 8: Launch and review weekly
Read through conversations every week. Look for questions it got wrong, questions it handed off that it could answer, and new questions to add to the knowledge base. Tell customers they are chatting with an automated assistant, and ask for their consent before collecting personal data.
How do you know the chatbot is working?
Track a few numbers from the first week, and compare them month to month:
- Answered without a handoff: the share of conversations it finished on its own.
- Wrong answers: how many you found in your weekly review. This number should fall.
- Leads captured: names and numbers collected that turned into a real conversation.
- Reply time: how quickly customers get a first answer, compared with before.
- Handoffs answered: how many handed-off customers a person replied to the same day.
If wrong answers stay high, fix the knowledge base before you add features. If handoffs pile up, your team needs time set aside to answer them.
Where chatbots help small businesses most
These are typical first chatbots for local businesses. They are examples of jobs a chatbot can do, not results we are promising.
Business | Chatbot job | Hand off to a person when |
|---|---|---|
Online clothing shop | Sizes, stock, delivery charges, return policy | A customer wants to change an order |
Clinic or salon | Timings, services, booking requests | Anything medical or a complaint |
Restaurant | Menu, delivery areas, opening hours | Large or catering orders |
Tutor or academy | Course details, schedules, fee questions | A parent asks about a specific student |
Real estate agent | Collect budget, area and phone number | The lead is ready for a viewing |
From chatbot to agent
A chatbot that only answers questions is step one. When it can also take actions, such as checking an order or booking a slot, it becomes an AI agent. What Is a No-Code AI Agent and How Do Businesses Use Them? explains that next step.
Building chatbots for local businesses is also a service freelancers can offer. SkillsVerx's AI Automations course covers chatbots, agents and the workflows behind them through real projects.
Frequently asked questions
How much does a small business chatbot cost?
It can start free on entry plans with low limits. Costs rise with conversations, channels such as WhatsApp, and AI usage. Estimate your monthly volume first.
Can a chatbot reply in Urdu?
Modern AI models handle Urdu and Roman Urdu reasonably well. Test replies carefully, and instruct it to reply in the language the customer uses.
Do I need a website to have a chatbot?
No. You can run it on WhatsApp through the WhatsApp Business Platform, or use a hosted chat page such as the one n8n provides.
Will the chatbot give wrong answers?
It can. Limit it to your knowledge base, forbid invented facts, add a human handoff and review conversations weekly.
Learn AI Automations hands-on
Turn this into a skill with a real, project-based course and a mentor in the room.