The conversational chatbot examples below are the exchanges that decide whether a visitor books a call or closes the window: the opening line, the qualifying question, the awkward price question and the hand-off. Each one shows a weak version, a stronger version and why the stronger one works, across the four industries we serve. At the end you will find how to turn the examples you like into instructions for an AI agent on the levelpartner.ai platform.
Example 1: the opening line, for an agency
Weak: "Hi there, how can I help you today?"
Stronger: "Looking at our paid search service? I can tell you whether it suits your budget in two questions."
The weak line asks the visitor to do the thinking. The stronger one names what they are looking at, offers a specific outcome and says how little effort it takes. Visitors reply to offers, not to open invitations.
Example 2: the qualifying question, for a consultancy
Weak: "Can you tell me about your business?"
Stronger: "Roughly how many people work in the business, and is the problem growth or capacity?"
A broad question gets a long answer or no answer. The stronger version asks two things the consultant genuinely needs, in a form someone can answer in a few words. It also signals that the firm understands the two common reasons people get in touch.
Example 3: the eligibility question, for a mortgage broker
Weak: "Do you qualify for a mortgage? Let me check."
Stronger: "I can't tell you what you'll be offered, but I can get you to the right adviser. Are you buying your first home, moving or remortgaging?"
The weak version implies advice the bot must not give. The stronger one sets the boundary honestly and moves straight to routing. Clients trust a bot more when it is clear about what it will not do.
Example 4: the price question, for real estate
Weak: "Our fees vary depending on the property."
Stronger: "Our fee depends on the sale price and the package you choose. A valuation is free and takes about 40 minutes, and you get the full fee in writing afterwards. Want me to find a time?"
Visitors ask about price because they want to know whether to keep talking. A vague answer ends the conversation. The stronger reply explains what drives the fee, gives a free next step and asks for the booking.
Example 5: the hand-off, any industry
Weak: "Let me transfer you to an agent."
Stronger: "That's a question for Sarah on our team. She's in until 5.30 today. Shall I ask her to message you here in the next hour, or would you prefer a call?"
The stronger hand-off names a person, gives a realistic time and offers a choice of channel. The visitor knows exactly what will happen next.
Before you start
- Access to AI Agents and the agent you want to improve.
- Ten recent transcripts from your current chat, or notes from phone enquiries.
- The examples above that match situations your firm meets.
How to set it up
- Go to AI Agents > Conversation AI and open Agent List.
- Select an existing bot or create a new one.
- Open the Bot Training tab.
- Choose an existing Knowledge Base from the selector, if you have one.
- Select All URLs with this Path to crawl one site section.
- Enter the full URL, including https://, and click Get Data.
- Review the pages found and select only the relevant ones.
- Click Train Bot and wait until each URL shows Trained.
- Click + Add Q & A and enter a question and answer.
- Save the Q&A and test variations of the question.
- Enter a question in Test your Bot and check the reply.
Why these work
Each stronger example does three things: it is specific to what the visitor is doing, it asks for a small next step, and it is honest about limits. None of them try to sound human. They sound competent, which matters more. A 10 to 250 person firm wins on answering the right question quickly, not on charm.
Suppose one chat in five currently goes past the first exchange. If replacing the greeting and the price answer with specific versions moved that to two in five, and bookings went from 9 to 16 a month on similar traffic, that would be seven extra calls a month from two rewritten lines. Track the same two numbers on your own chats before and after you change anything.
Copying example lines word for word into the agent makes every reply sound scripted. Give the agent the examples as a guide to style, and let it phrase replies to fit the conversation.
When not to script lines at all
If your agent handles mostly support questions from existing clients, these sales moments matter less. Focus on accurate answers and a clear hand-off instead.
Troubleshooting
The agent repeats the example line exactly. Tell it in the instructions that examples show tone, not words to repeat.
Replies get longer after adding examples. Add a length rule alongside the examples, such as two or three sentences.
The price answer still sounds evasive. Give the agent the real factors that set your price in the knowledge base.
Hand-offs promise times nobody can meet. Base the times on your business hours and on-call rota, not on a fixed promise.
Frequently asked questions
Can I use these examples on WhatsApp and SMS?
Yes, shortened. Messaging channels suit one-line replies. See building a WhatsApp chat bot for channel rules.
How many examples should an agent have?
Five to ten covering the moments that matter most. More than that and the instructions get hard to maintain.
Where do I find my weakest moments?
In transcripts where the visitor stopped replying. The AI customer service example article shows how to read them, and chat bot GPT covers instruction writing in depth.
Do these examples work for voice agents?
The principles do, but spoken replies need to be shorter still. The AI receptionist guide covers voice.