How to Think While Designing a Chatbot
This document defines the principles, mindset, and best practices that must be followed whenever creating or modifying a chatbot.
The goal is not to build an information dump.
The goal is to create a guided conversation that leads to an outcome.
A chatbot must feel like a conversation, not a FAQ page.
Information should never be dumped upfront.
The bot should:
Then provide the next relevant piece of information
Ask a question
Understand the response
Think like a human sales or support executive, not like documentation.
❌ Bad: Explaining everything in one message ✅ Good: Asking what the user wants, then responding step by step
Every chatbot flow should be interactive by design.
Ask questions frequently
Use buttons, quick replies, or choices wherever possible
Make the user feel involved in the conversation
If the user is only reading and not responding, the chatbot is failing.
Rule of thumb: After every 1–2 bot messages, the user should be encouraged to act.
Never send long paragraphs.
Never send dense text blocks.
Even if information is complex:
Break it into 2–3 short messages
Send them sequentially
Short sentences
One idea per message
Easy to scan on mobile
If it looks heavy on a phone screen, it’s wrong.
Whenever possible:
Use images, GIFs, or short videos
Especially for:
Product explanation
How something works
Social proof
Pricing or plans
Media should replace text, not accompany long text.
If an image can explain it, don’t write a paragraph.
The chatbot must continue the conversation the user already started elsewhere.
Always account for the source:
Ad click
Email campaign
WhatsApp broadcast
Website CTA
Support entry point
From an ad → acknowledge the ad intent
From an email → reference the email topic
From a campaign → continue that narrative
The user should never feel like they are starting from zero.
Whenever a human transfer is required:
Explicitly tell the user
Set expectations upfront
The bot should communicate:
Why a human is needed
What will happen next
When they can expect a response
Transparency builds trust. Silence breaks it.
Every chatbot must have a clear primary objective.
In most cases, this is:
Lead capture
Lead qualification
Name
Phone number
Email
Use-case / intent
Any qualification info relevant to the business
This should happen naturally through conversation, not as a form dump.
Try to auto-capture information wherever possible.
Never ask for everything at once.
Start light
Build trust
Ask deeper questions gradually
Example flow:
Understand intent
Provide relevant value
Ask for basic details
Qualify further if user is engaged
Avoid sounding salesy
Avoid forcing forms too early
Avoid repetitive follow-ups
Instead:
Offer help
Give choices
Let the user feel in control
A good chatbot guides. It does not pressure.
Always assume:
Users may give incomplete answers
Users may stop responding
Users may give unexpected inputs
Gracefully handle ambiguity
Gently re-prompt
Never blame the user
Never sound robotic when confused
Friendly
Warm
Simple language
No jargon unless user is clearly advanced
Avoid:
Corporate tone
Over-formal language
Long explanations
The chatbot should sound like:
“A smart, helpful person chatting on WhatsApp.”
Do not mix multiple objectives in one flow
If needed, branch intelligently
Keep the user oriented at all times
The user should always know:
What the bot is helping with
What comes next
When designing a chatbot:
Do not think message by message
Think:
What problem is the user solving?
What decision are they making?
What information do we need to move them forward?
Messages are just tools. Outcome is the goal.
Whenever creating or reviewing a chatbot, ask:
Is this conversational?
Is this engaging?
Is this easy to read on mobile?
Is this continuing the user’s context?
Is the end objective clear?
Are we capturing and qualifying leads naturally?
Would a human enjoy this conversation?
If the answer to any of these is no, redesign the flow.
A chatbot is not a knowledge base. It is a guided conversation designed to move a user from intent to action.
Last updated 1 month ago