Orchestration: the thread that holds a process together
Coordinates the steps of a process across systems that know nothing of each other
The process finishes even when one system falls over
Each step automated on its own works perfectly. The process breaks exactly between them, where nobody is looking.
What you get
One process, not five scripts
Steps are joined into one stateful flow, not a chain of automations each hoping about the next.
It retries by itself
A downed API does not stop the process. The step retries, and the rest waits instead of carrying on wrongly.
You see where it stopped
Every run leaves a visible history: which step, what data, what the other system replied.
Order matters and is kept
Nothing invoices before delivery is confirmed just because one message arrived sooner.
How it works
1. The trigger
An event starts the flow: an email arriving, a new row, a fixed time, or a call from another system.
- n8n
2. State
Each run keeps its own state, so it can resume from exactly where it stopped.
- PostgreSQL
3. The steps
Each step calls a system and checks the reply. Anything that answers wrongly does not pass onward.
- n8n
- REST
4. Retry
Temporary errors retry with growing pauses. Permanent ones stop and tell a person.
- n8n
5. Visibility
Every run is visible, with the duration and result of each step. Silence never counts as success.
- Slack
Where it works well
- Processes with steps across several systems
- Flows where order and state matter
- Integrations with unreliable or slow systems
- Processes that must resume where they left off
Where it does not
- A single step in a single system — a plain script fits better
- Processes that change weekly end up costing more than they save
- Decisions needing human judgement cannot be orchestrated, only routed to a person
What we use
- n8n
- PostgreSQL
Where we use it
Frequently asked questions
- How is this different from Zapier or Make?
- Those join two steps well. Orchestration holds a long process with state, resumption and guaranteed order — precisely what breaks as volume grows.
- What if the server dies mid-process?
- State is kept outside the process, so the run resumes from the last completed step rather than the beginning.
- Can we see what is happening, or only you?
- You can. The execution view is yours, with full access. It is not a dashboard we hold and describe to you.
- Is it self-hosted or does it depend on an external service?
- It can be hosted with you or with us. n8n is open source, so you are not stuck if you want to move it.
- How many flows can it handle?
- It needs no special infrastructure to run many flows in parallel. The practical limit is clarity — how many flows you can keep track of and explain — not capacity.