Who this is for. Super Admins. AP Helpdesk Managers, Specialists and Analysts cannot open AP Intelligence. See the Roles and Access.
The Rulebook holds rules written in plain language that guide how the agent handles mail. The tab describes itself as: “Manage organization-wide, team-level, and personal rules that guide the collaborative automation pipeline.”
It is the one place in the product where you tell the agent what to do in a sentence rather than by setting a field.
Figure 1. Rulebook Management.
Not to be confused with business rules. The Rulebook holds plain-language rules for how an AP Helpdesk agent handles mail. Business rules are IF-THEN logic that sets field values on an AP Invoices document. The two live in different places and do different work. Business rules are covered in the AP Invoices guide.
What a rule looks like
Rules are sentences, not forms:
Always escalate bank detail changes for manual verification
Escalate PO discrepancies over $1,000 to AP manager
Do not create email drafts for Invoice Submission email category
Assign all Inquiry and contact-update tasks to a named analyst
For all Update related tasks, set the Priority to Medium
They cover escalation, assignment, priority, and whether a draft is produced at all. That is the same ground as processing rules, expressed as instructions rather than settings.
Rulebook or processing rules?
Each row pairs the same decision from both sides. Choose the column that matches how the behavior is expressed and who needs to read it.
Processing rules are configured on the agent itself. See Create and Manage AP Helpdesk Processing Rules.
Table 1. Choosing where a behavior belongs
Use the Rulebook when |
Use processing rules when |
|---|---|
The behavior is a policy that reads naturally as a sentence. |
The behavior depends on an exact sender-type and category pair. |
It should apply regardless of sender type. |
It should apply only to one class of sender. |
You want it readable by someone who does not know the wizard. |
You need the precision of explicit conditions and actions. |
These rules run alongside the sender-aware processing rules configured on the agent, and neither overrides the other. If a behavior has no matching processing rule, look here. A rulebook entry is frequently the explanation.
Rule metadata
A Show rule metadata switch, at the top right of the rules list, adds an expand control to each rule row. Expanding a row shows the directive compiled from your wording, under Trigger, Action and Scope, which is useful when two rules appear to contradict each other and you need to see what each one resolved to.
Scope
Organization-wide, team-level and personal rules are listed together, so read the metadata with Show rule metadata when it matters which scope a rule carries. A personal rule affecting only its author looks identical in the list to one affecting everyone.
Add a rule to the Rulebook
Rules are added from Rulebook in the AP Intelligence left rail. A rule is a sentence, and the product checks it before it saves, so writing one is quick, but writing one that behaves takes a little care.
Add a rule
In the navigation rail, point to SmartVendor and select Settings from the menu that opens.
Confirm the instance selector names the agent you intend to change.
Open the Rulebook tab.
Write the rule as a plain sentence in the text box at the foot of the list, then select + Add Rule. This does not save the rule. It runs a safety and conflict check first.
Finish from the result: select Confirm & Add when the check is clean, or Add Anyway when potential conflicts are listed. A rule blocked for a safety or compliance violation cannot be added at all.
Figure 2. The entry box at the foot of the list, with Add Rule beneath it.
If you edit the text after a check, the button becomes Re-check Rule and you have to run the check again before you can save. The check applies to the exact wording, not to your intent.
Write a rule that behaves
The rules that work best share three habits.
Table 2. Three habits
Habit |
Example |
|---|---|
Name the trigger precisely |
“For external vendor emails containing ‘AR contact’… in the subject or body” |
State the action, not the intent |
“create a task for analyst acknowledgement” |
Say what happens by default |
“Default to no agent-drafted reply for Unknown sender” |
The second habit matters most. “Handle bank detail changes carefully” states an intent and gives the agent nothing to do; “Always escalate bank detail changes for manual verification” states an action and works.
Rules naming an individual, such as “assign to a named analyst”, break silently when that person leaves. Nothing in the product prompts a review, so a rulebook referencing people needs an owner and a periodic check.
Before you add one
Check whether an existing rule already covers it. Two rules on the same ground is how contradictions start.
Decide whether it belongs here or in processing rules. Sender-specific behavior usually belongs there.
Write it, run the check, and read any conflict the check reports rather than selecting Add Anyway reflexively.
Edit or remove a rule
Every rule in the list carries a toggle at the left of the row and its own menu at the right-hand end. The menu holds Edit, with Delete on your own personal rules.
Figure 3. The menu on a rule row.
Table 3. Maintaining a rule.
Control |
What it does |
|---|---|
Toggle |
Turns the rule off without deleting it. The rule stays in the list and stops affecting mail. This is the safer first move when a rule is misbehaving and you are not yet sure why. |
Edit |
Reopens the rule text. Saving runs the same conflict check that runs when you add one, so an edit that collides with another rule is caught before it lands. |
Delete |
Removes the rule permanently. It appears on your own personal rules. Organization and team rules can be edited and turned off from the row, but not deleted there. |
This is the half of the job to schedule rather than react to. A rule naming an individual keeps working after that person leaves. It continues assigning to that person, and the product raises no warning. Whoever owns the rulebook should read it through on a regular cadence and edit or turn off what no longer holds.
A rule still needs permission to act
A rule can be written correctly, pass its check and still do nothing visible, because the Rulebook decides what the agent should do and the pipeline switches decide whether it may act without a person.
A rule that drafts a reply produces a draft either way. Whether that draft is sent depends on Auto-reply. The same holds for forwarding and for document submission. If a rule appears to be ignored, check the switches before rewriting it. See AI Pipeline Settings.