Your automation stopped: call the original builder or hire someone new?

The short answer

Before you decide who fixes it, find out what kind of stop it is. That takes about 15 minutes and you can do it yourself: look at the run history, read the last error, check whether a login has expired and whether anything changed around the automation.

Then the choice is usually clear. If the cause is outside the automation (an expired login, a changed app, a limit) and the original builder answers, go back to them. They know the build. If they do not answer within the time you agreed, if the same failure keeps returning, or if the fix needs a change to how the whole thing works, bring in someone new, and hand them the workflow, the accounts and the log, not just “it stopped”.

First, find out what kind of stop it is

Most stops I have had in my own scheduled jobs were not bugs in the code. They were changes around it:

• A login stopped working. In September 2026 a token that one of my publishing jobs used to upload files was refused, and five publishing slots in a row over two days published nothing while every run reported success (the full incident).
• The machine was asleep. Scheduled jobs on a laptop ran hours late because the computer was in a low-power sleep mode, with nothing in the log to say so (that one too).
• A limit was reached. In September 2026 an overnight content job of mine hit a monthly spend limit at 01. Four slide decks had come out before it stopped, and the run ended with an error code instead of finishing.
• Something it depends on changed shape. When I changed the format of my posts, the publishing step still expected the old one in three places. A dry run showed it would have published the images and silently dropped the video. I found it before switching the schedule back on.

Each of these points to a different fix and a different person. Here is the check, in order:

Check	Where to look	What it tells you
Did it run at all?	The run history in Zapier, Make or n8n, or the scheduler	No runs: the trigger or the machine. Runs but nothing happens: a step inside.
What did the last run say?	The error on the failed step, or the log	“Unauthorized”, “401” or “reconnect”: a login. “Limit” or “quota”: an account limit.
Is every connected account still connected?	The platform’s connections or credentials page	A red or expired connection is a 10-minute fix for whoever has access.
Did anything change?	Your own memory, and the connected apps’ notices	A new column, a renamed field, a new plan, a moved server.
Is it every run or some runs?	The run history over a week	Every run: one cause. Some runs: data or timing, which takes longer to find.

Write down what you found. Whoever fixes it will ask these questions first, and having the answers turns an open-ended investigation into a short job.

When the original builder is the right call

Go back to the person who built it when:

• The cause is outside the automation and is what your maintenance terms call a fix: an expired login, an app that changed, a limit.
• They answer within the agreed time. If you agreed a response time in the maintenance terms, this is where it pays off.
• The automation has their logic in it. Rules, branches and AI prompts take time for a stranger to understand. The builder already does.

That has been my own habit. When I have had something built, the maintenance that follows has gone to the same developer, because they already know the build and the conversation starts in the middle rather than at the beginning. Fiverr reviews of automation sellers point the same way for many buyers: of 254 distinct buyers in the 317 reviews I analysed, 42 (17%) came back to the same seller.

When someone new is the right call

Bring in a new seller when:

• The builder does not reply within the time you agreed, or within a few working days if you agreed nothing.
• The same failure keeps coming back after each fix. That is a design problem, not a breakage, and a second pair of eyes is worth paying for.
• The fix is really a change. Your process moved on and the automation needs to follow. That is a new order anyway, so you can choose who takes it.
• The builder no longer works in that tool or quotes more for the fix than the build cost.

Switching has a cost. A new person has to read someone else’s workflow before touching it, and that reading time is part of the price. It is much smaller when you can hand over the things in the next section.

What to have in hand before you message anyone

Two messages you can copy

To the original builder:

Hi, the [name] automation you built stopped on [date]. The last run shows: [error]. The [account] connection shows [status]. Nothing changed on my side except [change / nothing I know of]. Is this covered by our fix terms, and when could you look at it?

To a new seller:

Hi, I have a [Zapier / Make / n8n] automation built by someone else that stopped on [date]. It [one-sentence goal]. I can give you access to the workflow in my account, the connected accounts and the error from the last run: [error]. Could you tell me whether this looks like a fix or a rebuild, and what you would need to quote?

A good new seller will ask to see the workflow before quoting, and will tell you if it looks like a rebuild. Among the sellers on the workflow automation page, Ramon C also offers live one-hour fix sessions, and the gig of Sam mentions maintenance for automations around custom or in-house systems. Both fit a stop that needs someone to look before they quote.

Next time: decide this before the first order

The best time to decide who fixes it is before you pay for the build. Agree a response time, a fix rate and a hand-over plan in the maintenance terms, and ask for an alert that tells you when the automation goes quiet, not only when it crashes (see heartbeat monitoring). And test the stops in advance: the acceptance checklist for unattended automations includes a run with an expired login for exactly this reason.

Then “it stopped” is a short message to someone who already agreed to answer it, not a search for a new freelancer while the work piles up.