A customer signs. Real money, a real commitment, the start of something you want to last years.
The first thing the company does is work out how to send them a message without a person involved.
Not whether the automation is good, whether the timing is right, whether the copy is warm enough. Look at the shape of the problem as it got framed. Somebody decided this was a throughput question.
I should say where I'm standing, because this reads like an anti-automation argument and it isn't one. I build agentic workflows for clients and for my own business, and I don't remember how I worked before them. That's precisely why I know what this particular one costs. You have to know what a machine can do before you can see clearly where putting one changes the thing itself.
The question you're actually answering
There are two, and teams answer the second one while believing they answered the first.
How fast can this go out is a throughput question. It has good answers. Same day, automatically, triggered on the closed-won flag, no human required.
What is this moment for is a different question, and it decides whether the first one matters.
If the answer is that you're starting a relationship, automation is a tool and a perfectly good one. If the answer is that you're discharging an obligation, then automation is the whole plan, and the relationship never actually starts. It just gets marked as having started.
The tell is what happens when speed and quality conflict. If the reflex is to protect the speed, you were answering the throughput question the entire time.
The guardrail
Automate something that's done right. Don't automate it for speed.
Speed is a property of a good process, not a reason for one. When it becomes the reason, you get systems that reliably do a thing nobody would have chosen to do by hand.
That's the actual risk, and it isn't new. Automating a broken motion doesn't fix the motion. It runs it more often, at less cost, with nobody in a position to notice. It's the same argument I made about three AI traps in customer experience: slowing down at the point where a person would have exercised judgment is what lets you go faster everywhere else.
What this doesn't mean
It doesn't mean send it slower. There's no virtue in a customer waiting three days for a welcome.
It doesn't mean write every message yourself. If a hundred customers a week need the same first email, and it's a good email, send it a hundred times.
The claim is narrower and harder to wriggle out of. Decide what the moment is for before you decide how it gets delivered. Design it as though a person were doing it well, then automate the version that survives.
The first message after somebody pays you is the cheapest, highest-leverage thing you will ever send. Almost nobody treats it as a design problem, because it arrives on a list of things to operationalize rather than a list of things to get right.
