- Original event
- Retain the class-change request and its identity
- Approval
- Confirm the new time; a refusal follows a separate path
- Reminder
- Stop the old reminder and check the relevant participants
- Message
- Prepare a new message; sending, delivery and reading remain distinct
- Exception
- Handle timeout, retry and cancellation without claiming an unchecked success
What people and teams gain
Routine transitions follow a consistent path.
The event, rule and permitted action are known.
A retry does not create a second effect.
The same event is checked before another change is made.
A person can see exceptions.
The process can stop rather than silently creating an error.
The result can be verified.
Before reporting success, we check whether the required action was completed.
An event becomes an action after checks and approval
Class or appointment changes
Receive the agreed event and retain its identity.
At event receipt: repeated event
Check the existing outcome before another action.
Check schedule, resources and confirmation
Rules may be sufficient; AI is optional.
At the check: refusal, unavailable resource or timeout
Keep the exception visible for the responsible person.
Perform and verify the permitted change
A proposal becomes a booking only after required checks.
After the accepted change: old reminder
Cancel or replace the outdated reminder in the agreed scenario.
Example: a proposed appointment becomes a booking only after resource availability and customer confirmation. A refusal, timeout or repeated event follows its own branch; ordinary rules do not automatically require AI.
Start with the rule, then choose the tool
Useful automation starts with an event: a request arrives, a document is approved or a status changes. We need to understand which information is checked, who decides and what should actually happen. An unclear manual process does not become reliable simply by being put into a diagram.
AI is optional. If a defined rule solves the task, we use it; we add a model only for a useful step whose result can be checked.
What we organise
The event and required data
We define the source, identifier and starting conditions. We check that enough information is available and that the event has not already been processed. An incomplete request receives a visible status and an owner instead of being lost between systems.
Actions and approvals
We assemble the selected steps: a check, task, notification, document preparation or permitted change. Human decisions remain where accepting terms, professional judgment or financial authorisation is required. If approval is refused, the action is not performed. The person responsible can see the refusal status.
A retry without a second effect
We agree how the result is confirmed and when it is safe to retry a step. No response does not prove that an action was not performed. If the outcome is unknown, we first check the state or pass the question to the person responsible.
Exceptions and stopping
We account for service outages, incorrect data, cancellation and changes to the original event. The process needs a clear way to stop, continue after a check or finish without further action.
Class change: check each transition
Two useful scenarios
An enquiry to a service company. The system checks the necessary details, assigns an owner and prepares a task. The employee sees the original request and the next step. Automation does not send an unapproved commercial offer on their behalf.
How pNet works
1
Choose a recurring event. Define the source, data and result of one specific transition.
2
Agree the rule and roles. Decide who authorises the action and what happens when data is missing.
3
Set up execution and retry. Check operation uniqueness, errors and safe stopping.
4
Check the complete result. Verify the recipient's state as well as the request that was sent.
Real messages and changes are enabled only after agreed checks and permission. We provide instructions for monitoring and exceptions. After launch, we help investigate errors, refine rules and maintain the interfaces in use.
What the team receives
More than a visual diagram: a clear working process showing which steps are complete, what is awaiting approval, where an error occurred and who needs to intervene. Logs contain the necessary details without unnecessary secrets or personal data.
For a Beauty salon, an available time slot can be offered to someone on a waiting list, but an offer does not become a booking until the resource and customer confirm it. A supplier can use an approved quotation to trigger the next step, but not automatic dispatch without stock and an authorised order. An academy does not assign places to individual participants merely because a corporate package was purchased.
We describe the event, condition, owner, status and exception. Simple transitions use rules; AI can help prepare content checks when needed. A retry after a timeout must not create a duplicate booking, charge or message.
Events and data come from CRM/Odoo and verified integrations; monitoring shows when a process has stopped and what can happen next.
Important: Automation does not have to use AI: a normal rule with a check is often simpler and more reliable.
Frequently asked questions
Yes. The process can prepare a draft or task, with a person approving the action. This is a useful way to check rules and data quality.
No. A retry may create a duplicate or a second payment. Each step needs clear conditions and a check of the previous result.
No. A deterministic rule is usually easier to check when inputs and conditions are defined.
The rules cover this: stopping the old action, obtaining new approval or making a correction. A cancelled booking must affect a reminder that has not yet been sent.
Connected services
Choose one recurring path
Tell us what starts the process, who is involved and which steps are repeated manually today. We will define a useful scope and safe boundaries.