Who Fixes the Robot When It Breaks? Name an Owner

A customer mentions, almost in passing, that they never got the appointment reminder this time. You nod, say sorry, and make a note to look into it later. Then you check, and the reminder hasn't gone out to anyone in about three weeks.

Nobody turned it off on purpose. A password expired, or a setting got bumped during an app update, and the little robot that texts your customers the day before just stopped. No error landed in your inbox. No one complained either, because people rarely call to say "I didn't get a text." They just show up late, or not at all.

Here's the uncomfortable part. Automation is great at doing the same thing every day without being asked. It's bad at telling you when it has quietly stopped. Something has to run unattended, and someone has to own the fact that it's running, and those are two different jobs. The second one is the job most shops forget to hand to anybody.

How robots break, and why you won't notice

Most automation doesn't fail with a bang. It fails with a shrug. Knowing the common ways it goes dark tells you where to look.

• A step gets paused. Someone clicks the wrong toggle during an update, or pauses a sequence so it won't text over a holiday and forgets to switch it back on.
• An integration drops. An integration is just the handshake that lets two apps talk, your booking tool handing a phone number to your text tool, and that handshake breaks when one side changes.
• A login expires. Passwords and access tokens (the digital key one app uses to log into another) time out on a schedule, and when the key dies, the robot can't get in.
• An input changes. You rename a field on your intake form from "Phone" to "Mobile," and the robot that was looking for "Phone" now finds nothing.
• A limit gets hit. Starter plans cap how much you can send, and once you cross the line the sends just stop until the month resets.

None of these throw a flare. That's the whole problem, and it's why "I'll notice if something's wrong" is not a plan.

The weekly check: five minutes with your coffee

The cheapest insurance is a short look, once a week, same time every week. Put it on your calendar like you would a standing delivery. The goal isn't to audit everything, it's to confirm the robots you depend on actually did their job.

Keep the check boringly simple. For each automation you rely on, answer one question: did it run this week, and did it run for the right number of people?

1. Pull up the activity log or sent history for each key automation. Most tools keep one.
2. Check the count against reality. If you had twelve jobs booked and only four reminders went out, something's off.
3. Send yourself a test where you can. Book a fake appointment with your own number and watch the text land.
4. Glance at anything sitting in an error or "failed" column. That's the robot waving a small flag.
5. Write down the date you checked, so next week you know the window you're responsible for.

The alert that pages a human

That weekly check catches slow leaks. It won't help much if the robot dies on a Monday and your whole week leans on it. On the automations that really matter, set up an alert so the system tells you the moment it fails, instead of you hearing about it first from a customer.

Two kinds of alerts earn their keep.

• Failure alerts. When a step errors out, the tool emails or texts a real person. Most platforms can do this, you just have to turn it on.
• Heartbeat alerts. If a robot that normally fires every day hasn't done a thing in 24 hours, that silence is itself the warning.

Be honest about what an alert can and can't do. It can tell you the robot threw an error or went silent. It can't tell you the robot did the wrong thing with full confidence, like sending a polished message to the wrong customer. Catching that still takes a human eye, which is why the weekly check and the alert work together rather than one replacing the other.

Naming the owner: you, your office manager, or someone you hire

Every automation needs a name attached to it. Not a department, not "the system," a person who'd feel a little embarrassed if it broke and sat broken for a month. That person has three realistic candidates.

You, the owner

Fine when you've got one or two simple robots and you're close to the work. The risk is that you're also the busiest person on the crew, and the weekly check is exactly the kind of quiet task that loses to a loud day.

Your office manager or a key staffer

Usually the best fit once you're past a couple of automations. They sit at a desk, they see the customer side, and they'll hear about a dead reminder before you do. Give them the authority to pause, test, and ask for help, and write down what "normal" looks like so they're not guessing.

Someone you hire

For a bigger stack of automations, or ones wired straight into money and scheduling, it's fair to pay the person or company who built them to keep watch. You're buying their judgement about what broke and how to fix it, not just a warm body. Make sure "watch it" is written into the deal, because "we built it" and "we're keeping it alive" are not the same arrangement.

Whoever you pick, the split stays the same. The machine runs the relay, day after day, without complaint. The person owns the judgement calls: is this normal, is this broken, does this need a fix today or can it wait until Friday.

Keep a one-page map

When something breaks at the worst possible time, the owner shouldn't have to reverse-engineer the whole setup from scratch. Keep a single page, a document or even a pinned note, that lists each automation in plain English.

List the job it does, the apps it connects, where the on/off switch lives, and who to call if it's past a quick fix. Takes an afternoon to write once. Saves the frantic hour later when you're trying to remember which login runs the thing.

Worth doing this week

You don't need to overhaul anything. Pick the robot you'd miss most if it vanished tomorrow, and start there.

• Write down your top three automations and what each one actually does.
• Put a recurring five-minute check on your calendar, same time weekly, and do the first one today.
• Turn on failure alerts for anything tied to money, booking, or customer messages.
• Name an owner for each robot out loud, and tell that person it's theirs.
• Start the one-page map, even if you only fill in a single line today.

We build these relays for a living, and the first question we ask is never "what can we automate," it's "who notices when it stops." Get that answer right and the rest holds up. The robots do the repetitive relay. You and your crew keep the judgement, which was always the part worth keeping.