Customer service scripts and examples for phone and chat
Customer service scripts for phone and chat calls: greetings, apologies, escalations, and tech support triage, with real data on when scripts backfire.
Is it right for you?
- Write greetings, apologies, and escalation lines as starting points, not verbatim scripts.
- Build separate chat templates instead of copying phone scripts word for word; chat needs shorter sentences and a faster first reply.
- Ask what changed and what the customer already tried before starting a troubleshooting script.
- Give agents a documented way to skip or shorten a script when it does not fit the situation.
- Review actual call and chat transcripts monthly and update scripts based on what agents are already saying that works.
Quick verdict
Use scripts as checklists, not lines to recite. Fixed word-for-word wording earns its place for mandatory disclosures on compliance-heavy calls; everything else in a small support team goes faster and lands better as a flexible template an agent can adapt mid-sentence. Sample calls monthly, let agents help rewrite the lines that keep getting skipped, and move a scenario into decision-tree or AI-prompting software only once volume makes manual scripting unreliable.
What a customer service script should actually do
A good customer service script is a checklist an agent glances at, not a monologue read word for word. It should guarantee a few things happen every time: a real greeting, the customer's issue confirmed back to them in plain language, an apology that names the actual problem, and a clear next step before the call or chat ends. The templates below cover the scenarios small support teams handle daily: opening a call, apologizing for a mistake, escalating a complaint, triaging a technical issue, and running a live chat on a tighter clock than a phone call allows.
Reading from a script tends to backfire. A Software Advice customer survey, summarized by CallMiner, found that 69% of respondents said their experience improved noticeably once an agent stopped sounding like they were reading from a script [Software Advice survey via CallMiner, accessed 2026-08-18]. Balto's Conversation Excellence Lab, which surveyed 567 contact center agents, found two-thirds want to change the scripts they were handed, and that agents who help write their own scripts follow them more closely and report higher job satisfaction than agents who are simply handed one from above [Balto Conversation Excellence Lab, "The Case for the Anti-Script," accessed 2026-08-18].
"We call Balto the 'anti-script.' The anti-script doesn't mean agents should be winging it on every call. Rather, it means giving agents the ability to be human and stay flexible when talking to customers."
Marc Bernstein, CEO of Balto, on the Conversation Excellence Lab's script-adherence research
None of that makes scripts pointless. Compliance-heavy calls in insurance, collections, healthcare, and financial services still need mandatory disclosures read the same way every time, and that is a different problem with a different tool; see our call center scripting software comparison for the enforcement-grade tools built for that job. Everything below is for the far more common case: a small support team that wants consistent quality without turning agents into voice actors.
Phone greeting, hold, and transfer templates
Opening greeting
Thank you for calling [Business Name], this is [Agent Name]. How can I help you today?
Confirming the issue
Just so I have this right, you're calling about [restate the issue in one sentence]. Is that accurate?
Placing on hold
I need about [realistic time estimate] to [specific reason: pull up your account, check with billing]. Can I put you on a brief hold?
Returning from hold
Thanks for waiting. Here's what I found: [answer].
Transferring
I'm going to connect you with [name or department], who handles [specific reason]. I'll give them a quick summary first so you don't have to repeat everything.
Closing
Before we finish, is there anything else I can help with today? You should [next step and timeline]. Thanks for calling [Business Name].
Two rules make these hold up across a real shift. First, never place someone on hold without a stated reason and a time estimate, "one moment please" reads as a stall even when the wait is genuinely short. Second, the transfer line only works if the agent actually passes along the summary; a customer who has to repeat the whole story to a second person remembers that more than anything said in the first three minutes of the call.
Apologizing without sounding like a form letter
The widely used LAST framework, Listen, Apologize, Solve, Thank, breaks service recovery into four steps: let the customer finish explaining the problem without interrupting, apologize for the specific impact rather than a stock phrase, fix the issue or explain the concrete plan to fix it, and thank the customer for raising it [LAST service-recovery framework, as described by Customer Service Manager and CenterEdge Software, accessed 2026-08-18].
Minor issue (late reply, small billing error): I'm sorry that happened, that's on us. Here's what I'm doing to fix it right now: [specific action]. And here's how I'm making sure it doesn't happen again on this account: [if applicable].
Repeated or serious issue: I want to acknowledge this wasn't a one-time slip, and I understand why that's frustrating to deal with twice. Here's what actually happened: [brief, honest explanation]. Here's what I'm doing right now: [action]. And here's who to contact directly if [specific follow-up scenario].
Skip any version of "we're sorry for any inconvenience this may have caused." That phrase could apply to literally anything, which is exactly why it reads as insincere; naming the actual thing that went wrong is what signals a real person read the ticket. Balto's agent-satisfaction research points at the same gap from the other side: you can coach an agent on what to say, but empathy that sounds scripted reads as fake no matter how the words are arranged [Balto Conversation Excellence Lab, accessed 2026-08-18].
Escalation and complaint scripts
| Trigger | What the agent says | Next action |
|---|---|---|
| Customer asks for a manager | "I can get my supervisor involved. Before I do, can you tell me what resolution you're hoping for so I can pass that along and we don't lose time repeating the story?" | Brief the supervisor with the stated ask, not just "please escalate." |
| Customer threatens to cancel or leave a bad review | "I hear you, and I don't want you to leave unhappy. Let me see what I can actually do here." | Check the account for real options (credit, plan change) instead of promising something outside your authority. |
| Second or third contact about the same issue | "I can see this is the second time you've had to reach out about this. I'm not going to make you go through the basic steps again." | Skip standard troubleshooting and route straight to a root-cause owner. |
| Customer becomes abusive and won't de-escalate | "I want to help you, and I can do that better if we keep this respectful. If [specific behavior] continues, I'll need to end the call and have someone follow up in writing." | Follow the company's written de-escalation policy rather than individual judgment alone. |
The pattern across all four rows: acknowledge what the customer actually said before moving to the next step. An agent who jumps straight to "let me transfer you" without confirming the ask forces the customer to explain everything twice, once to the first agent and again to whoever picks up next.
Tech support triage: ask before you troubleshoot
CompTIA's standard troubleshooting methodology puts information-gathering before any fix attempt: identify the symptoms, ask what changed, and confirm whether the problem is isolated to one user or affecting others, before touching a single setting [CompTIA troubleshooting methodology, accessed 2026-08-18]. Support scripts that skip straight to "have you tried restarting" waste time whenever the customer already tried the obvious fix, or when the real cause is a recent change rather than a broken device.
Opening triage
Before we start, three quick questions so I don't have you repeat anything you've already tried: What were you doing right before this started? Have you already tried [the most common self-fix for this issue]? And can you give me the exact error message, or send a screenshot?
If they already tried the common fix
Got it, thanks for trying that first. Since that didn't fix it, let's go one level deeper: [next diagnostic step specific to the product].
If nothing has changed on their end
That's useful, it means this is more likely something on our side or a recent update. Let me check [system status/recent release notes] before I have you change any settings.
Closing a troubleshooting session
Here's what fixed it, and here's what to try first if it happens again: [specific recap]. I'm also logging this so if you contact us again about the same issue, whoever picks up can see exactly what we already tried.
That last line matters as much as the fix itself. A customer who has to re-explain a resolved issue from scratch on a second contact is the single most common trigger for the "second contact about the same issue" escalation row above.
Chat scripts need a different clock than phone scripts
Chat has no tone of voice to soften a delay, so silence reads as rudeness faster than it would on a phone call. Zendesk's own guidance on first reply time cites 40 seconds or less as a strong benchmark for live chat, drawing on research from Toister Performance Solutions, Sprout Social, and Call Centre Helper [Zendesk, "First reply time," accessed 2026-08-18]. That gap between a 40-second target and how long a full answer actually takes to research is why a chat greeting needs to fire the moment a conversation opens, even before the real answer is ready.
Instant acknowledgment (send before the full answer is ready): Hi, thanks for reaching out. I'm looking into this now, one moment.
Greeting with context (when a pre-chat form captured the topic): Hi [Name], I see you're asking about [topic]. Let me pull that up.
Delay update (if it's taking longer than expected): Still checking on this, I'd rather get you the right answer than a fast wrong one. Should have it in about [time estimate].
Escalating chat to a call or email
This one will be faster to sort out over a quick call or a detailed email so I can [specific reason: share a screen, attach a file]. Are you free for [specific option]?
Closing a chat
Does that cover everything, or is there something else I can look at while we're here?
Keep chat sentences shorter than their phone equivalents. A customer reading a wall of text on a support widget interprets length as a stall the same way a caller interprets an unexplained hold.
Fixed script, decision tree, or AI real-time prompting
Which tool fits depends on team size and how much of the volume is repeat-scenario. A small team handling low call volume does fine with a shared document like the templates above, memorized loosely and adapted per conversation. A growing team that needs the same disclosure or troubleshooting branch triggered consistently benefits from decision-tree software, where the next line depends on what the customer just said. We compared the two leading approaches to that problem in Zingtree vs Balto: Zingtree is the click-through decision-tree model, while Balto listens to the live call and surfaces prompts and compliance alerts without the agent navigating anything manually. If you're evaluating the wider market, we tested seven scripting and AI-coaching tools on compliance depth and AI guidance.
If most of your team's scripted conversations are outbound sales calls rather than inbound support, the flow is different enough to need a separate playbook built around booking a next step instead of resolving a problem; see our outbound sales call script examples for cold-call, referral, and voicemail templates. And if a meaningful share of your inbound volume never reaches a live agent at all, the fix might not be a script: a clearly written voicemail greeting that sets an accurate callback window handles a surprising share of after-hours contacts on its own.
Reviewing scripts without treating them as gospel
Sample real calls and chats monthly rather than scoring every interaction against the exact wording. Look specifically for two things: where agents consistently diverge from the script, and whether that divergence is a good adaptation worth keeping or a sign the script no longer matches how customers actually describe the problem. If you're already recording call outcomes for attribution, the same data helps here; pull disposition codes and recordings the way our call tracking metrics guide describes, and look at calls where handle time spiked, that's usually where the script stopped fitting the real conversation.
Involve agents in rewriting the lines that keep getting skipped. Balto's research on script adherence found this is not a soft management gesture, agents who help write their scripts follow them more closely and report meaningfully higher job satisfaction than agents working from a script written entirely by someone else [Balto Conversation Excellence Lab, accessed 2026-08-18].
Frequently asked questions
Q: Should agents read a customer service script word for word?
A: Generally no, outside of compliance-mandated disclosures. A Software Advice survey summarized by CallMiner found 69% of respondents said their experience improved when agents stopped sounding scripted [Software Advice survey via CallMiner, accessed 2026-08-18].
Q: What is the LAST method in customer service?
A: A four-step service recovery framework: Listen, Apologize, Solve, Thank. It structures how an agent responds to a complaint without requiring fixed wording [Customer Service Manager; CenterEdge Software, accessed 2026-08-18].
Q: How is a chat script different from a phone script?
A: Chat needs shorter sentences and a much faster first response, since there's no tone of voice to soften a delay. Zendesk's own guidance on first reply time treats 40 seconds or less as a strong benchmark for live chat [Zendesk, "First reply time," accessed 2026-08-18].
Q: When should a support team move from a written script to scripting software?
A: Once the team is large enough, or the compliance requirements strict enough, that a shared document can't reliably guarantee the right branch or disclosure fires every time. Our guide to that software category covers what it looks like at scale.
Q: How often should customer service scripts be updated?
A: Review real transcripts monthly rather than on a fixed annual cycle, and update lines that agents are already skipping or rewriting on their own in live conversations.
Q: What's the biggest mistake teams make with apology scripts?
A: Using a generic line like "sorry for any inconvenience" that could apply to any situation. Naming the specific problem is what makes an apology read as genuine rather than templated.