Managed automation is business automation that keeps being operated after it goes live: someone monitors the runs, gets an alert when something fails, and fixes and updates it against the systems it depends on. Instead of a workflow that is handed over and forgotten, it is a standing operations layer over processes in n8n, WhatsApp, a CRM and spreadsheets.
The reason it matters is simple. An automation is not a file written once. It is a live link between several systems, and each of them changes without asking anyone.
What is the difference between managed automation and automation that is handed over and left?
Managed automation has an ongoing operational owner: someone who monitors the runs, receives an alert when something fails and deals with it. Automation that is handed over and left is a workflow that was built, tested and delivered, after which responsibility for failures passes to the business. Both can work well on handover day. The difference shows a few weeks later.
In a handed-over project the question “who fixes it?” stays open. In practice the answer is whoever notices first, and that is usually a customer who got no reply or a salesperson who never received a lead. With managed automation the answer is settled in advance: someone sees the failure before the customer does, and there is a clear procedure for what happens next.
Why does an automation stop working when nobody touched it?
An automation stops working even when nobody touched it because it depends on external systems that change: an access token that expired, a CRM field that was renamed, a spreadsheet column that moved, or a WhatsApp message template that was paused. The workflow itself stays the same, but the environment around it does not.
A few concrete examples:
- Tokens. A temporary token from Meta’s developer console expires after a short time. If it is still in use in production, messages simply stop going out.
- The 24-hour window in WhatsApp. With the WhatsApp Cloud API you can send free-form text only within 24 hours of the customer’s last message. Outside that window you need a pre-approved template. A process built without this in mind will pass testing and fail in real use.
- Spreadsheets. Someone inserts a column in the middle of the table, and the process starts reading the wrong field. There is no error, only wrong data.
The last case is the most dangerous, because nothing “breaks” in a way anyone can see.
What happens when an automation fails and nobody notices?
When an automation fails and nobody is alerted, the damage builds up quietly: leads do not reach the CRM, messages are not sent, invoices are not issued, and the business finds out only when a customer complains. In n8n every run is stored with a status, and you can set up a dedicated error workflow. But that only helps if someone receives the alert and acts on it.
In n8n you can also set a node to retry after a failure (Retry On Fail). That solves momentary faults, such as a server that did not answer the first time. It does not solve a real fault, such as a revoked permission or a missing required field. Those need a person who understands the process and knows where the data was supposed to go.
What does managed automation include in practice?
Managed automation includes five standing components: run monitoring, error handling, updates against external systems, planned changes to the process, and regular reporting to the business owner. The exact scope depends on each business’s processes and systems, but the table shows the skeleton.
| Component | What is done in practice | Example |
|---|---|---|
| Monitoring | Checking that processes run as planned, including processes that should run at a fixed time | A workflow that should run every hour and did not triggers an alert |
| Error handling | Classifying each failure: retry, data fix or system fault | A lead without a phone number is held in a review queue instead of being lost |
| Updates against systems | Adjusting the process when an external system changes | A CRM field got a new name, and the mapping is updated |
| Changes and improvements | Adding a step, condition or connection, tested before it goes to production | Adding a WhatsApp reminder to an existing process |
| Reporting | Showing the owner what ran, what failed and what was fixed | A periodic summary of errors and fixes |
What is not included: business decisions. If the business changes its sales process or its pricing, that is a change to the process and not maintenance, and it needs to be discussed.
How do you monitor an automation without flooding everyone with alerts?
You monitor an automation well by sorting failures into types and sending each type to a different channel: momentary faults are handled by an automatic retry, data problems go into a queue for review, and a system-level fault raises an immediate alert. An alert for every single error teaches people to ignore all of them.
There is also a kind of failure with no error message: a process that did not run at all. So absence of activity is monitored too. A process that should handle leads every hour and handled nothing beyond the set time is a sign of a fault, even if nothing reported a failure.
Who holds the access, the code and the data?
The accounts, API keys and data belong to the business, even when someone else operates the automation. That is how any sound managed arrangement should work: access is registered in the business’s name, and keys are kept in the platform’s credential store rather than inside process nodes. It also protects you: if you part ways with a provider, the processes and accounts stay with the business.
We bring working habits from banking environments. Each connection gets only the permissions it needs. Changes are tested before they go to production. What changed, and when, is documented. And testing is kept separate from the process that handles real customers.
Is managed automation right for every business?
Managed automation fits a business whose automated process touches customers, leads or money, meaning wherever a silent failure is expensive. It is less necessary for a small internal process where any failure is immediately visible, such as a one-off data export to a spreadsheet.
A few questions help you decide:
- Could a customer or a lead be harmed if the process stopped for a few hours?
- Does the process depend on three or more systems, such as WhatsApp, a CRM and an invoicing system like Morning?
- Is there someone in the business who can read run history and fix a workflow?
If the answer to the last question is no and the first two are yes, that is exactly where outside management saves you surprises. The same principle applies to Make and Zapier: every platform has run history, and every platform has external systems that change.
How do you get started with managed automation?
You start with a review of the existing or planned processes: which systems are connected, which data moves between them, and what the damage is if a process stops. On that basis you decide what to monitor, who receives alerts and what response times apply. You can begin with one process and expand.
If you do not have a first automation yet, start with the guide on launching a first automation. To see how we build and operate processes like these, read about our managed automation service.
Frequently asked questions
Can a workflow someone else built be moved under management?
Usually yes. The process starts by reading the existing workflows, checking the access and connections, and adding error handling and alerts where they are missing. Only then do you decide what to monitor and what to improve. If there is no documentation, it is written as part of the review.
How is managed automation different from ordinary technical support?
Technical support responds when someone reports a problem. In managed automation the monitoring runs in advance, so a fault is identified before a customer or employee notices it. Management also includes updating the processes when external systems change, not only fixing them after a failure.
How is managed automation priced?
Pricing is usually based on the number of processes, the complexity of the connections and the service level required, for example the response time to an alert. That is why a fair price cannot be set without seeing the actual processes. We do not publish a flat price here, and we define the scope after a review.
Do I need managed automation if I use a simple platform like Zapier?
Yes, when the process touches customers or money. A simple platform makes building easier, but it does not remove the dependence on external systems that change. There too, someone needs to check the run history and act when something fails.