The work that only happens because somebody remembers.

One of them, or the whole chain.

Each job above is a chain. Take one link and build it on its own, and that is the small price. Build the whole chain, with the branches and the places it has to stop, and that is the bigger one. Nothing here is priced on how many people you employ.

What happens when it breaks, in two honest halves.

This is the question everybody asks last and worries about first. The honest answer has two parts, and only one of them is easy.

Both run without anyone watching, at any hour. A person replies during working hours — we would rather write that down than promise somebody is awake at three in the morning and be caught out the first time we are not.

Most of what a broken job costs is not the repair. It is the gap before anyone knows. The fix is usually short once somebody is looking at it. What is long is the stretch between the thing failing and anyone finding out, and that stretch runs on whether the job happens to be noticed. We have no measured figure to put on it and we are not going to borrow one. That gap is what the monthly removes.

And most of what breaks is not yours. It is a supplier changing something at their end — a field renamed, a login method replaced, an old version switched off. A study of 1,368 public APIs found that when one of these breaking changes ships, close to nine in ten arrive with no deprecation notice at all (Yasmin, Tian and Yang, ICSME 2020: 219 of 251 versions, 87.3%). That is nine in ten of the changes that break something, not nine in ten services. So we watch what your suppliers announce as well as what your own system does, because the alternative is finding out from a customer.

It is built in your account, never in ours.

You are the owner and the billing contact from the first day. That sounds like paperwork until the day you want to leave.

• On Zapier this is not reversible.

  Zapier states plainly that an account cannot be transferred to somebody else. If an agency builds your automations inside its own Zapier account, they are theirs, and the only way out is to build the whole thing again somewhere else.

• Nothing stops working if you stop paying us.

  No licence, no seat count, no key we hold. Hand the whole thing to your own developer or to another firm whenever you like, and we will help them pick it up.

• You get a written guide, not a handover call.

  What each part does, how to change the common things, and what to do when something looks wrong — written for the person who will actually run it.

The useful part is what it refuses to do.

• It never decides anything that costs money.

  It can work out that an invoice looks wrong, or that a delivery is short. A person still says what happens next. Every system on this site draws that line in the same place.

• It stops, and it says why.

  A job that cannot finish does not fail quietly. It stops in front of a named person with the reason attached, which is the whole difference between something that helps and something you have to check.

• We will not automate a job nobody can explain.

  A job done four different ways by four different people is not one job yet. Running it on a machine does not settle which way is right — it just picks one and makes it permanent, at speed, without telling anyone.

• We will not sell you a build you do not need.

  If the job is twenty minutes of work, we will say so and lose the fee. A job we talk you out of costs us one invoice. A job that should never have been built costs us the reference, and we do not have many.

The promise, with the conditions next to it.

• A written list of checks, agreed before we start.

  We write down what the finished thing has to do, and you agree it before anything is built. That list is the only test. If a check fails we keep fixing it at no extra cost until it passes, however many times that takes.

• A date, and a clock that pauses.

  You get a date in writing when the list is agreed. The clock pauses while we are waiting on you — logins, answers, a decision. We tell you the day it pauses and the day it starts again.

• We come back on day 7, day 14 and day 30.

  Named days, not “thirty days of support”. The first week is when a job meets real data and behaves differently; the second is when the person using it has opinions; the thirtieth is when we find out whether it is still being used at all.

• A full refund inside seven days if it misses the scope.

  The written list above is the scope. If what we deliver does not do what that list says, email us within seven days of delivery and the payment comes back. That is the same clause in our terms, not a friendlier version of it for a sales page.