A good AI voice agent opening line is short, names who is calling and why, says it is an AI, and ends with one easy question. Aim for about 15 seconds. This guide gives five steps and 10 example lines.
Step 1: Pick one job for the first line
Give the opening one job: earn the next 20 seconds. It does not have to finish the task. A first line that tries to explain the offer, verify identity and collect a payment will lose the listener before it ends.
Start with your campaign type. Insight7's voice campaigns, which are in beta, are set up as a customer survey, lead follow-up, cold outreach or win-back. Each type has a different first job:
- Survey: get permission for a short question.
- Lead follow-up: remind them of the request they made.
- Cold outreach: give one believable reason for the call.
- Win-back: show you noticed they left.
People hang up on unknown numbers because they can't tell who is calling or why. They also hang up on scripted-sounding voices, and buyers tell us this fear is common. So the first line must sound like a person with a purpose.
Step 2: Say who, why and that it's AI
Put your company, your reason and the AI disclosure in the first sentence. A listener decides in seconds whether a call is a scam. A line like "Hi, I'm an AI assistant calling for [Company] about your recent signup" covers all three.
Disclosure rules differ by region. Here are three to check:
- EU: Article 50 of the AI Act says people must be told they are dealing with an AI, unless that is obvious. The notice must come no later than the first interaction. Cooley's guide confirms it has applied since 2 August 2026.
- US: AI-voice calls already need consent under the TCPA, following the FCC's 2024 ruling. A rule requiring "say it is AI at the start" was still not final in mid-2026.
- Colorado: The Chatbot Safety Act takes effect on 1 January 2027. It requires any public-facing conversational AI to disclose it is AI.
Check the law where your customers live.
Offer the human route early. In a Gartner survey of 3,566 customers, 87% said companies using GenAI must offer a way to reach a human. So say "you can ask for a person at any time" early.
Prepare the "where did you get my number" answer. Callers will ask, so state how the contact got on the list, such as "You gave us this number when you signed up." Insight7's voice campaigns, in beta, include consent marking. Check in the app how it works for your campaign. If you have no true answer, remove that contact.
Step 3: Give them a reason to stay
Open with a detail only this customer would know, then ask one easy question. The detail can be the signup, the due date or the install. A pitch gives the listener nothing to recognise, so they hang up.
Use two or three short sentences:
- Say what you know: "You signed up on Monday."
- Say why you are calling: "Your ID check isn't finished."
- Ask a yes/no or choice question: "Do you have two minutes now, or would later today be better?"
Never end on "How can I help you?" It hands the work back to the customer and invites the agent to drift. Before an agent talks to real customers, check for three faults: long openings, drifting and looping.
Step 4: Sound human
Write for the ear. Use contractions like "I'm" and "you're". Use the customer's name only if your data is reliable, because a wrong name ends the call faster than no name. Read the line aloud. If you run out of breath, cut it.
Match the voice to the listener. Buyers ask for local accents and languages, because a local voice matters for trust. In Insight7 Voice, which is in beta, a manager picks the campaign language before the voice, and Yoruba and Igbo are available. Nigerian Pidgin is also a new campaign language, in beta.
Copy these 10 opening lines
Each line names the AI, gives a reason and ends on one easy question. Swap the placeholders for your own details. Read each one aloud first.
-
Customer survey: "Hi [Name], I'm an AI assistant calling for [Company]. You spoke with our team last week, and we'd like your feedback. Do you have two minutes now?"
Why it works: it asks permission first. -
Lead follow-up: "Hi [Name], this is [Company]'s AI assistant. You asked about [Product] on our site yesterday. Is now a good time for two quick questions?"
Why it works: the customer recognises their own request. -
Cold outreach: "Hello, I'm an AI assistant calling for [Company]. We help [role] teams with [one problem]. Is that part of your work, or should I ask for someone else?"
Why it works: it gives one believable reason and an easy way out. -
Win-back: "Hi [Name], I'm [Company]'s AI assistant. We noticed you haven't used [Product] since [month]. Would you tell us why, in about a minute?"
Why it works: it shows you noticed, and asks for little. -
Pre-due-date payment reminder: "Hi [Name], I'm an AI assistant for [Company]. Your payment is due on [date]. Would you like a quick reminder of the amount?"
Why it works: reminders work best when they feel helpful, not like collections. -
KYC or onboarding nudge: "Hi [Name], I'm [Company]'s AI assistant. Your ID check isn't finished yet. Can we complete it now, or would later today be better?"
Why it works: it offers a choice and targets a common drop-off point, where customers start KYC and never finish. -
Post-install satisfaction check: "Hi [Name], I'm an AI assistant from [Company]. We installed your [product] on [date]. Is everything working as you expected?"
Why it works: it checks in before a complaint arrives, which buyers outside banking want. -
Appointment confirmation: "Hi [Name], I'm [Company]'s AI assistant. You have an appointment on [day] at [time]. Can you still make it?"
Why it works: it needs only a yes or no. -
Service outage or delay notice: "Hello [Name], I'm an AI assistant for [Company]. Your [service] is delayed today. Would you like the new time, or a person to call you?"
Why it works: it gives bad news first and offers the human route at once. -
Candidate screening invite: "Hi [Name], I'm [Company]'s AI assistant. You applied for the [role] job. Do you have ten minutes now for a short voice screening?"
Why it works: it names the application and the time. HR and onboarding teams are asking for this.
For line 10, Insight7's candidate screening, which is in beta, can run as a live voice conversation. Each candidate gets a private link.
Adapt each line to your campaign type, disclosure rules and customer language before you launch.
Step 5: Test the line before launch
Test the opening line in four passes before it reaches a real customer. Start in the browser. In Insight7's beta, a campaign's agent can be tested by talking to it there, with no phone number needed. Then place a test call to your own phone, which the campaign setup also offers in beta. Listen on a real handset, because a phone line sounds different from laptop speakers.
Next, check the edges. Write the voicemail wording on purpose. Set calling hours for when your customers are free. Campaigns have calling hours in beta.
Then launch to a small batch and measure the first seconds. Read each transcript and note where people hang up or say "who is this?" Change one variable per round, such as the first sentence, the voice or the closing question. Otherwise you will not know what helped.
Insight7 puts these steps in one place. The voice features are in beta. A dashboard shows calls in progress and lets a reviewer read each transcript when the call ends. A campaign also gets a scorecard when it ends.
Related reading: How AI Voice Agents Handle Interruptions, Background Noise and Cross-Talk, How to Fix AI Voice Agent Mispronunciation of Customer Names, Brand Names and Acronyms and 11 Reasons Customers Hang Up on AI Voice Calls and How to Fix Each One.
Questions people also ask
How long should an AI voice agent opening line be?
Two or three short sentences, about 15 seconds when read aloud. That is our own guideline. A longer opening gives the listener time to hang up before your question arrives.
Should an AI voice agent say it is an AI at the start of a call?
Yes. Say it in the first sentence. The EU requires disclosure now, and Colorado will from 2027. A customer who works it out later feels tricked, and that ends the call faster than an honest opening.
What should an AI voice agent say when someone asks where it got their number?
Name the source and offer removal: "You gave us this number when you applied. I can take you off the list now if you'd like." Insight7's voice campaigns, in beta, leave do-not-contact people out of copied campaigns.
How do I stop customers hanging up on an AI voice call?
Cut the length, open with a detail only that customer knows, and ask a yes/no question. Call at sensible hours. Local numbers are more likely to be answered, and Insight7's beta supports Nigerian numbers.
How do I make an AI voice agent sound less robotic?
Write for the ear: short sentences, contractions, one idea each. Spell out product names the way a customer would say them. Then listen to real recordings and rewrite any line that sounds read aloud.
How do I test an AI voice agent before it calls real customers?
Talk to the agent in the browser first, which Insight7 allows with no phone number, in beta. Then call your own phone. Try interrupting it and staying silent. Then run a small batch.
What should an AI voice agent say when the call goes to voicemail?
Keep it to about ten seconds: who is calling, that it is an AI, one reason and a callback route. Leave out personal details.
Listen to calls, then refine
A strong opening line is short, honest and specific. Review real calls and fix the first 15 seconds before anything else. In your next meeting, ask: in the first 15 seconds, what does the customer learn, and what must they do?

