## TL;DR
Canned responses work for the repeatable 80 percent: status updates, how-tos, policy explanations, closing messages. Personalization is required for the human 20 percent: the first sentence that proves you read the ticket, anything emotional, and anything unusual. The line is simple: template the structure and the facts, personalize the acknowledgement and the judgment calls. A macro with one genuine sentence beats a fully hand-written reply that says nothing.

## The query

```text
canned responses vs personalization: where the line is
```

## Use this when

- Building or auditing a macro library
- Replies sound robotic and CSAT is dipping
- Agents spend too long hand-writing repeatable answers
- You need a rule for when to go off-script

## Not for

- Chatbot conversation scripting
- Marketing or lifecycle email copy
- AI support agent prompts
- Legal or compliance responses

## The line

### Always safe to template

Status updates, how-to instructions, policy explanations, password and billing procedures, closing messages, follow-up bumps. The facts dont change, so the words shouldnt either.

### Always personalize

The first sentence of any reply, responses to angry or upset customers, anything unusual or edge-case, apologies, and any no. These are the moments the customer is checking whether a human is there.

### The hybrid: template with a personal slot

Most macros should be 80 percent template, 20 percent fill-in. The template carries the structure and facts, the agent writes one or two sentences of genuine acknowledgement.

## Steps

### 1. Audit your top 20 macros for the personal slot

Open each macro and mark where the agent adds something human. If a macro has no slot for it, add one, usually the first sentence.

Expected output: every macro has a marked personalization slot.

### 2. Write macros with brackets, not blanks

"[One sentence naming their specific problem]. Here is how to fix it: ..." Brackets tell the agent exactly what to write and where.

Expected output: macros that guide the personal touch instead of hoping for it.

### 3. Ban the phrases that signal a macro

"Per our policy," "kindly," "we appreciate your patience" used alone. Replace with plain language that says the same thing.

Expected output: a short banned-phrase list the team actually follows.

### 4. Let agents go off-script for the emotional 20 percent

Explicitly tell the team: if the customer is upset, confused, or the situation is weird, drop the macro and write it fresh. Permission matters, agents follow the template when they are unsure.

Expected output: a written rule that off-script is allowed and when.

### 5. Review macro usage monthly

Check which macros get heavy edits. A macro agents rewrite every time is a bad macro, fix it or kill it. A macro nobody touches might be perfect, or might be sending robotic replies nobody flagged.

Expected output: a monthly macro review with edit-rate data.

## Variant phrasings

### when to use canned responses in customer support

Use them for the repeatable 80 percent listed above. The test: if the facts are identical across 50 tickets, template it.

### how to personalize canned responses

The personal slot in the first sentence. Name their problem in your own words, then let the template carry the rest.

### support macros that dont sound like templates

Brackets for the human part, banned phrases out, plain language in. The macro should read like a good agent on a good day.

## Why it happens

Teams swing between two failure modes: everything hand-written (slow, inconsistent) or everything templated (fast, robotic). The line exists because customers dont actually mind templates, they mind feeling unheard. One genuine sentence at the top proves a human read the ticket, and then the template's efficiency is a feature, not an insult.

## Edge cases

- High-volume simple tickets: full templates are fine, the personal slot can be a one-clause insert ("Got it, your export failed again").
- Angry customers: never lead with a macro. Write the first two sentences fresh, then you can use templated steps for the fix.
- Non-native English agents: macros are a gift, they guarantee correct grammar. Put the personal slot in simple language they can write confidently.
- AI-assisted replies: same rule. Let the AI draft the template parts, require the human to write or verify the acknowledgement sentence.

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_0iLY0KtzMeYD9u85B6dTCw