This morning I published a post about missed call text back, the automation that answers a caller before a human can. It felt like the right day to admit how I handle my own follow-up, because it is not what you might expect from someone who sells automation: half of mine is deliberately not automatic.
The weekend just gone
Over the weekend, a routine read my inbox every couple of hours. Each pass it did the same small jobs: filed the notifications and receipts that did not need me, kept the threads that did, and for one conversation where a reply was owed, wrote out the reply it proposed I send.
It proposed that reply on Saturday morning. It proposed it again at lunch, and again in the evening, and through Sunday, and in the small hours of Monday morning. The same thread, the same careful draft, pass after pass.
It never sent it. It is not allowed to.
There was also an email address that bounced. The routine did not quietly retry it or guess a correction. It flagged it and left it, because deciding what to do about a dead address is my job, not its.
Why the machine drafts and I send
That design is not timidity about automation. I automate sends all day: the post this morning is literally about a message that goes out in seconds with nobody touching it.
The difference is what kind of follow-up it is. A missed call text back has a known moment and a known message. The trigger is fixed, the wording is approved in advance, and the right response time is measured in seconds. Automating that is not risky, it is the only way to be fast enough.
A reply to a real conversation is the opposite. The moment is known but the message is not: it carries tone, judgement about what to offer, and sometimes the decision not to reply at all. So the routine does the part that is mechanical, reading everything and preparing the draft, and leaves the part that is judgement. When I sit down this morning, the thinking is already done and waiting. Reading a prepared draft and deciding takes a minute. Starting from a blank reply takes a morning.
The rule I ended up with
Automate the moment, keep the judgement. If the moment and the message are both known in advance, the send should be automatic, because speed is the whole value. If the message needs a human decision, automate everything up to the send, and never the send itself.
Nothing went out in my name over the weekend. And the reply was ready before the working day started. Both of those are the system working.
That split is exactly what I set up for clients on NexBrand: the fixed-moment messages run themselves, and the judgement replies arrive as drafts in a shared inbox with the history attached. The missed call piece is in this morning's post.
Frequently Asked Questions
Why not let the system send the replies too?
Because those replies carry judgment: tone, timing, what to concede and what to ask. The routine drafts so the thinking is done, and I send so the judgment stays mine. That split is deliberate.
What does the routine actually do?
Every couple of hours it reads the inbox, files the noise, proposes a reply where one is owed, and flags anything it should not decide. It sends nothing. The drafts wait for me.
When should a follow-up be fully automatic?
When the moment and the message are both known in advance: a missed call, an appointment reminder, a quote that has gone quiet. Fixed moment, approved message, automatic send. Everything else gets a human.