Every support team has the macro that agents quietly stopped using. It opens with "Thank you for contacting us regarding your inquiry," it answers a slightly different question than the one the customer asked, and it ends with a line nobody would say aloud. Agents retype around it, or send it unedited and watch the customer write back, "Did you read my message?"
Canned responses are worth having. They keep answers consistent across a team and across shifts, and they save the typing. The trick is writing them so the reply still reads as an answer to this person. These are the habits that tend to work.

Write blocks, not whole replies
A full-message template assumes every customer asks the same thing in the same order. A block assumes only that a certain piece of information is needed. Split your library into three kinds of pieces:
- Openers that acknowledge the specific situation: "Sorry the invite link didn't work."
- Answer blocks that hold the policy or steps, written once and kept accurate.
- Closers that say what happens next and who does it.
An agent assembles a reply from two or three blocks and adds a sentence of their own. The answer block carries the consistency. The opener and the added sentence carry the human part. When a policy changes, you edit one block instead of forty macros.
Leave visible gaps for the specifics
The macro that sounds canned is usually the one that lets an agent send it without looking. Build in prompts that force a look:
Example textHi {first name}, thanks for the details about [what they were trying to do]. That error usually means [cause, in one sentence]. Here's how to fix it:
- ...
- ...
If that doesn't work, reply with [the one thing we need next] and I'll take it from there.
Bracketed fields that must be replaced before sending do two jobs. They make the reply specific, and they make it hard to send the wrong template, because the agent has to read the customer's message to fill them. Name the fields by what goes in them, not "[insert here]."
Match how your team actually talks
Read ten recent replies from your best agents. Note the length, the greeting, whether they use contractions, how they apologize. Then write macros in that voice. A few guidelines that hold up across teams:
- Lead with the answer or the next step, not with gratitude for writing in.
- Use plain words. "Sign in" beats "authenticate," unless the customer used the other one.
- Apologize once, for the actual problem, and only when something went wrong on your side. A reflexive "we apologize for any inconvenience" on every reply reads as a form letter.
- Keep sentences short enough to read on a phone.
- Say who is doing what: "I've passed this to our billing team" is different from "this has been escalated," and customers can tell.
Different channels want different lengths. A chat reply of a few lines would look curt in email, and an email-length block pasted into chat is a wall of text. Keep a version of each frequently used block per channel, or write blocks short and let agents extend them in email.
Say what you cannot promise
Macros are where overpromising spreads, because one hopeful sentence gets sent hundreds of times. Review every block for phrases such as "right away," "within 24 hours," "this will be fixed," or "you'll hear from us shortly." Keep a commitment only if the team really meets it. If the timing depends on something, say what it depends on. Where a customer's case needs an exception, a decision or a policy owner, the macro should say a person will review it, and not imply the outcome.
This is also the place to check what is not yours to say. If a block touches a legal, security or contract matter, have the person who owns that area approve the wording, and note that approval and the date in the macro's description.
Give every macro an owner and a review date
Libraries rot. A macro written for last year's product tells customers to click a button that moved. Keep a short record for each block: who owns it, when it was last checked, and which policy or article it depends on. Then:
- Review the most-used blocks monthly and the rest quarterly.
- After any product or policy change, search the library for the old wording and fix it the same day.
- Retire macros nobody has used in a quarter, instead of letting the list grow past what an agent can search.
- Let agents flag a block that did not fit. Their edits are the best evidence of what to rewrite. If many agents rewrite the same sentence, that sentence is wrong.
A simple signal to watch: for each macro, how often the customer's next message is "that's not what I asked" or the same question again. A high rate means the block answers the wrong thing.
Test a macro before it goes in the library
Before a new block ships, have two agents who did not write it answer three real recent tickets with it. Ask them where they had to rewrite, where they hesitated, and whether the bracketed fields made sense. A block that needs heavy editing in the first test will need it every time. Also read it once as a customer would, with no knowledge of your product's vocabulary. Internal names for plans, queues or features often sneak into macros, and customers do not know them.
When not to use a canned response
Some messages should not get one. An upset customer who describes a specific, painful situation wants to see that someone read it. A message that mixes two issues needs a reply that addresses both, in the customer's order. And a first reply to a long, detailed message should never be a short template. Teach agents to recognize these, and tell them that choosing to write from scratch is not a failure of efficiency. It is the right use of their time.
Where RyteDesk fits
If you use an AI assistant as part of your front line, the same rules apply to what it says: answers should come from approved knowledge, in your team's voice, with clear handoffs. RyteDesk is an AI customer support assistant in early access, for mid-market teams. It answers on chat, email and phone only from the knowledge you provide, you can try it in a preview that sends nothing, and you can read every conversation to see where the wording needs work. Join the waitlist if you would like to try it; it is free to join.