Automation should save the team's attention, not create invisible behaviour that people are afraid to touch. In Apexloop, automations are node-based flows. Each flow has exactly one trigger and then a series of steps that work with context.
This shape matters: when it is clear what starts the flow and what follows, the process can be read, debugged and extended.
Choose one clear trigger
A trigger can be a manual button, a schedule, a webhook, a received email or notification, or a CRUD change on a record. Good flows start with a specific event:
- user clicked "Issue invoice",
- every day at 8:00 overdue records are checked,
- a webhook arrived from an external service,
- a new record was created from a form,
- a deal status changed to "Won".
If you feel a flow needs multiple triggers, you usually have multiple processes. Split them.
Pass context forward
Each node can pass context to its successors. This allows working with data the flow has already found or created. Typical nodes are CRUD record, find records, iterator, condition, send message, HTTP and webhook.
Example billing flow:
- Button on a project starts the flow.
- Flow finds the customer and project line items.
- Condition checks that billing details are not missing.
- CRUD node creates the invoice and line items.
- Document template is filled with the data.
- Flow sends an email or saves the output to the record.
The user sees one click, but the process remains readable.
Only change non-computed fields
Automation can change non-computed values via an update node — that is, fields without a formula. Computed fields should remain under the control of formulas. This prevents the flow from manually overwriting values that should be calculated from data.
In practice: automation can set status, responsible person, deadline or a link to a created record. Sums of line items, margins or risk flags should be left to formulas.
Automation and permissions
For sensitive processes, think about column-level restrictions. Prices, personal data, approval fields or internal notes don't need to be visible to everyone. A flow should respect the data model and store results where the team expects them.
If a record is archived and other columns are read-only, it is good to have clearly defined which automations are still allowed to change it and why.
Integrations belong at the edges of the flow
Webhook and HTTP nodes connect Apexloop to external services. Manage secrets via connectors — personal or shared as needed. If an integration requires special logic, it is possible to prepare a custom node.
Try to keep the main business process in Apexloop and external calls as clearly named steps. When an integration changes, you don't have to rewrite the entire workflow.
A good automation test
Before deploying, walk through three scenarios:
- the normal case, where all data exists,
- a missing required value,
- a record that has already been processed or archived.
If the flow shows a clear result in each scenario, the team will start to trust it. Automation then is not a black box. It is a documented process.