Skip to main content
Procedures instruct the AI “how to answer the customer,” not what internal actions to perform. Each procedure follows the pattern: “When a customer asks … if … then …”

Procedure Patterns

PatternDescriptionWhen to Use
Single If-Then (Binary Rule)One condition, one responseSimple, clear-cut scenarios
Multi-Tree RuleSeveral if branches beneath one triggerMultiple possible conditions/outcomes
Workflow-Style RuleStep-by-step guidance with checks in sequenceTroubleshooting or multi-step support flows
Templated Response RulePicks a ready-made reply template based on conditionsFast, consistent messaging with variables

Key Points

  1. Focus on the reply—describe what the AI should “say” or whether it should “escalate” to human staff
  2. Stay conditional—use clear “if” clauses so the AI knows when to apply rules
  3. Keep each rule self-contained—no need to reference other Procedures

Where to Find Procedures

Navigate to: Agent Hub > Wiki > Procedures Define conditions and instructions that guide AI responses based on message type, contact details, or other triggers. All Procedures apply only to the selected inbox.

Procedure Templates

Conduit offers preset templates for common procedures by industry. Users can fully customize templates in their workspace. To find Procedure templates:
  1. Navigate to Agent Hub > Knowledge > Wiki > Procedures > Add Procedure
  2. Select “Import template” and choose from Conduit’s predefined procedures

Good vs Bad Procedures

Example 1 - Condition Clarity

Bad: “If needed, extend the deadline”—AI doesn’t understand “needed” Good: “If a customer requests a deadline extension and no conflicting commitments exist, approve up to 48 hours”

Example 2 - Customer-Focused Replies

Bad: “If customer reports an issue, notify operations manager and create task”—AI cannot perform back-office actions Good: “If customer reports an issue, thank them and confirm the team has been informed”

Example 3 - Conditional Statements

Bad: “Offer 20% discount for repeat customers”—no context for when to use Good: “If customer mentions prior purchases, offer 20% discount on next order”

Example 4 - Knowledge vs Procedures

Bad in Procedures: “We don’t offer refunds”—static policy, belongs in Knowledge Good: “If customer requests a refund within 30 days, process it; otherwise explain the policy”

Example 5 - Guardrails vs Procedures

Vague: “Do not disclose pricing differences”—permanent restriction Procedure: “If customer asks why price differs elsewhere, explain platform variations” Guardrail: “Never mention platform price differences in communications”

Example 6 - Outcome Clarity

Bad: “If product is broken, fix it”—no customer-facing instruction Good: “If customer reports a defect, ask for details and order number, then confirm the team will follow up”

Example 7 - Verifiable Conditions

Bad: “If customer requests expedited shipping and warehouse confirms stock, ship it”—AI can’t verify real-time inventory Good: “If customer requests expedited shipping, confirm you’ll check availability and notify them with an update”

How the Reasoning Engine Uses Your Procedures

The processing steps are:
  1. Engine reads all Procedures and matches “if” conditions against contact data and chat context
  2. Performs time/price calculations (e.g., “hours until arrival,” “percentage refund”)
  3. Drafts reply following chosen policy, applies Style, and passes through Guardrails before sending

FAQ

What information can I refer to in Procedures?

Reference contact details (account status, payment status, key dates), chat history, Knowledge base, and calendar availability. More data sources coming soon.

When should I use templated replies?

Use carefully and only when exact wording is mandatory. These responses are sent verbatim, requiring very precise “if conditions” to prevent misuse. Key considerations:
  • Use only when specific wording is legally/policy-required
  • Include detailed context for when template should/shouldn’t be used
  • Provide clear “if conditions” that must be met
  • Consider adding fallback options
Example well-defined templated rule: “If customer requests VAT invoice AND order is completed AND paid in full, use template: ‘Thank you for requesting a VAT invoice. I’ll prepare it within 24 hours. Please confirm company details: company name, address, VAT number.’”

How do I test that the AI follows a Procedure?

Open Sandbox, adjust contact details in Settings, impersonate the customer, send the triggering question, and verify the reply.