Chatbot conversation design: writing bots people like

How to design chatbot personalities, greetings, flows, error handling and tone, so AI assistants feel helpful, clear and on-brand.

By the UK Technology Hub teamPublished 4 minute read

The short answer

Good conversation design makes a chatbot feel clear, helpful and consistent with your brand. Define a simple persona and tone, greet users with what the bot can help with, ask one question at a time, offer buttons for common choices while allowing free text, confirm before taking actions, recover gracefully from misunderstandings and end conversations with a clear outcome.

AI models can generate fluent replies, but fluent isn't the same as helpful. Conversation design shapes how an assistant introduces itself, guides users, handles confusion and represents your brand.

Define the persona

A simple persona keeps tone consistent:

  • Name (optional) and clear disclosure that it's an AI assistant.
  • Personality traits: friendly, calm, concise, professional.
  • Voice: formal or casual, matching your brand.
  • Boundaries: what it helps with and what it doesn't.

Avoid overly human personas that might mislead users.

Greetings that set expectations

Tell users what the assistant can do: "Hi, I'm the Acme assistant, an AI that can help with orders, returns and product questions. What can I help with?" Offer a few common options as buttons.

One question at a time

Ask a single, clear question per message, especially when collecting information. Long multi-part questions confuse users.

Buttons and free text

Buttons speed up common paths and reduce errors. Free text handles everything else. Use both: suggestions for typical choices, plus the ability to type.

Keep replies concise

Short paragraphs, bullet points for steps and links to detailed pages. Long walls of text overwhelm chat windows, especially on mobile.

Clarify when unsure

If a question is ambiguous, ask a clarifying question rather than guessing: "Do you mean delivery times for the UK or international orders?"

Confirm before actions

Before booking, cancelling or changing anything, summarise and ask for confirmation.

Graceful error recovery

When the bot doesn't understand or can't help:

  • Admit it plainly.
  • Offer alternatives or rephrasing tips.
  • Offer a person. Our handoff guide explains how.

Avoid repeating "Sorry, I didn't understand" loops.

Sensitive moments

Complaints, bereavement, distress and safety concerns need careful wording and quick escalation. Write and test responses for these scenarios.

Ending conversations

Close with a clear outcome: what was done, any next steps and an invitation to ask anything else. Optionally ask for a quick rating.

Tone across channels

Website chat, WhatsApp and voice may need slightly different styles. Voice replies should be even shorter.

Grounding and accuracy

Design determines how answers feel; knowledge determines whether they're right. Our chatbot knowledge base guide covers content.

Testing with real users

Watch people use the assistant, read conversations and refine wording, flows and buttons based on where they struggle.

Common mistakes

  • Pretending to be human.
  • Long, rambling replies.
  • No way out of loops.
  • Inconsistent tone.
  • Taking actions without confirmation.

A worked example

A gym chain's chatbot asked new visitors for their name, email, phone number, fitness goals and preferred location before answering any question. Most people left. The team redesigned the flow so the assistant answered questions first, offered membership options based on one or two simple questions and asked for contact details only when visitors wanted to book a tour. Completion rates improved significantly, and visitors rated the experience more highly.

Writing sample conversations

Before building, write sample conversations for common scenarios, including when things go wrong. Read them aloud with colleagues to check they sound natural, helpful and on brand.

Designing for accessibility

  • Keep messages short and clear.
  • Make buttons accessible to keyboard and screen reader users.
  • Avoid relying only on images, emojis or colour.
  • Give users enough time to read and respond.

Personality without distraction

A friendly personality helps, but avoid jokes or chatty replies that slow people down. Personality should support clarity, not get in the way.

Iterating

Conversation design isn't finished at launch. Review real conversations, find where people struggle and refine wording, flows and options regularly.

Handling off-topic questions

Visitors will ask unrelated questions. Politely explain what the assistant can help with, offer relevant options and avoid inventing answers outside its scope. A friendly redirect keeps the conversation useful without frustrating the person.

Designing for returning users

Returning customers shouldn't have to repeat information. Where they're logged in and have agreed, the assistant can remember recent orders or bookings and offer relevant help straight away.

Small wording changes often make a big difference.

How we help

Our AI chatbot development team designs conversation flows and personas, our design team aligns tone and interface with your brand and our AI solutions team tests and refines assistants with real conversations.

Frequently asked questions

Should my chatbot have a name?

It can, as long as it's clearly presented as an AI assistant and not mistaken for a person.

Are buttons better than free text?

Use both. Buttons speed common paths; free text handles everything else.

How long should chatbot replies be?

Short and scannable, with links for more detail.

This guide is general information, not legal, tax or financial advice, and details such as prices, features and rules can change. Product and company names mentioned are trademarks of their respective owners and are used only to identify those products; UK Technology Hub is not affiliated with or endorsed by them.

Related guides

All AI chatbots & assistants guides