Prepared ANET – Creating Intents & Conditional Workflows for Non-Emergency Calls
Audience: Call Takers, Dispatchers, Supervisors, Administrators
Support Contact: Support@prepared911.com
Introduction
Welcome to the ANET Non-Emergency Training Packet for Creating Intents & Conditional Workflows. This course will guide you step-by-step through building Intents for common non-emergency calls in the ANET system using the Think Backwards approach. You will learn how to navigate the system, build and edit intents, and route calls efficiently.
By the end, you will be able to:
- Identify and correctly classify common non-emergency calls
- Apply the Think Backwards method to intent building
- Use system tools to ask the right caller questions
- Avoid misroutes and unnecessary escalations
1. Understanding Intents in ANET
An intent represents the purpose of a caller’s interaction. For non-emergency workflows, intents allow AI to:
● Identify the caller’s need.
● Ask targeted follow-up questions.
● Route or resolve the call using predefined conditions.
Example:
Intent name: Downed Powerlines
Description: Identify reports of powerlines fallen onto the roadway.
The Think Backwards Approach
In this method, you start by identifying the final action the call requires—where it will end up—and then work backwards to determine the path. This ensures every caller is routed accurately and quickly.
Example: If a caller reports an abandoned vehicle on the street, the final action may be to transfer them to the Police Non-Emergency line. Working backwards, you determine the right questions to confirm location, vehicle details, and whether the vehicle poses a hazard.
Your goal: Always know the end before you begin. This reduces errors and improves caller satisfaction.
Before you start building anything in ANET, you need to understand how to design your workflows for the best caller experience.
We use the Think Backwards approach to do this.
Why It Works
When you’re building call workflows in ANET, it’s tempting to start by listing every question you might want to ask. But if you don’t know exactly where the call needs to go in the end, you risk:
● Asking unnecessary questions slows down the interaction.
● Creating overlapping conditions that confuse the AI.
● Sending calls to the wrong department because the decision points weren’t clear.
By starting with the end action — like “Transfer to Police Non-Emergency” or “Provide Animal Services link,” — you keep every question focused on reaching that outcome as quickly and accurately as possible.
How to Apply It in ANET:
Start with the Outcome: Decide where the call needs to go (e.g., “Transfer to Model City 311” or “Send community resource link”).
Identify the Decision Points: Ask yourself what you must know to confirm this is the right action.
Write Targeted Questions: Only include questions that directly affect routing or are essential for the call notes.
Test the Shortest Path: Make sure the caller can reach the correct outcome with the fewest possible questions.
Example – Animal in Vehicle:
● Final Action: Transfer to Police Non-Emergency.
● Key Decision Points:
○ Is the animal in danger?
○ Is the vehicle unattended?
● Backwards-Built Questions:
○ “Where is the vehicle located?”
○ “Is the engine running or windows open?”
○ “How long has the animal been inside?”
When you build this way, you create logical, caller-friendly, and easy-to-maintain workflows.
How Parent and Child Intents Work
In ANET, you start with a Parent Intent. This is the main category of calls you’re building, such as “Abandoned Vehicle” or “Animal Services.”
Within that parent, you’ll create Child Conditions — these are the specific situations that trigger different actions.
Think of it like this:
Parent: Abandoned Vehicle
Child Condition: Vehicle in driveway
Child Condition: Vehicle on private property
Child Condition: Vehicle on public street
Each child condition will have its own caller questions and routing action. For example, look at the following scenario. Below you will find examples of parent intents and child conditions.
Scenario 1 – Graffiti Complaint
Purpose: Handle calls about graffiti quickly and determine correct routing.
Caller Questions:
● Is the graffiti currently in progress, or has it already been completed?
● Where is the graffiti located?
● Is it on public property, private property, or a business?
● Does it contain hate symbols, threats, or gang markings?
● When did you first notice it?
● Do you see the person responsible right now?
● Can you describe the suspect or their vehicle?
● Can you provide your name and callback number?
● Conditional Routing Paths:
● In progress → Police Non-Emergency
● Already done, no suspect → Code Enforcement
● Contains threats/hate speech → Police Non-Emergency
Example Parent Intent: Abandoned Vehicle
Description: For calls reporting abandoned vehicles in different locations.
Caller Information:
● Can I have your name and callback number?
● What is the exact location of the vehicle?
Child Condition 1: Driveway
Criteria: Vehicle is in a residential driveway.
Caller Questions:
● Is the vehicle blocking access to the driveway?
● How long has the vehicle been there?
● Can you describe the make, model, color, and any visible license plate?
● Do you know the owner of the vehicle?
● Has anyone attempted to move it?
Action: Transfer to Model City 311 (###-###-####).
Child Condition 2: Private Property
Criteria: Vehicle is on private property not owned by the caller.
Caller Questions:
● Is the vehicle creating a hazard or blocking access?
● How long has the vehicle been there?
● Can you describe the make, model, color, and any visible license plate?
● Do you know the owner of the property?
● Has the property owner been contacted?
Action: Advise caller to contact a private towing company; transfer to Police Non-Emergency (###-###-####).
Child Condition 3: Public Street
Criteria: Vehicle is on a public street or highway.
Caller Questions:
● Is the vehicle creating a traffic hazard or blocking a lane?
● How long has the vehicle been there?
● Can you describe the make, model, color, and any visible license plate?
● Does the vehicle appear damaged or vandalized?
● Do you know who owns the vehicle?
Action: Transfer to Police Non-Emergency (###-###-####).
Create a Parent Intent and Child Conditions (Step‑by‑Step)
BEFORE YOU BEGIN:
Collect your transfer numbers, business hours, and any self‑service links or canned messages you plan to use.
1) Go to “Settings” →” Non-Emergency”
Select “Intents” tab.
Click on “New Intent” to the right.
Enter a clear, descriptive name (for example, “Abandoned Vehicle”). Use action‑oriented language when possible.
Enter a one‑sentence description that states when to use this intent and what outcome it drives.
In Caller Information, add the questions the AI should always ask first (for example, name, callback, exact location).
Enter the questions that the AI agent should always ask the caller.
Determine if this will be a single path or a conditional path. Single paths are Intents that have a constant outcome. Whereas, the conditional path's response will vary by the conditions you set forth.
Set the Child Conditions. You can choose a single condition or several conditions.
CAUTION:
Avoid conditions that can be true at the same time. Overlapping conditions cause unpredictable routing.
Example: Time‑of‑Day Injection
You can route differently by time of day. For example, for downed powerlines:
- Condition A: “Sparking AND before 10 a.m.” → Forward to FIRE AM
- Condition B: “Sparking AND on/after 10 a.m.” → Forward to FIRE PM
- Condition C: “Not sparking AND traffic not blocked” → Provide utility info & request photo
Determine what action should be taken.
Save the intent.
Add Transfers, Canned Messages, and Toggles
For each child condition, set the Transfer Immediate Number and contact label (for example, “Police Non‑Emergency”).
Add Canned Messages for self‑service or follow‑up (for example, a link to Animal Services reporting or a photo request instruction).
Use the ON/OFF Toggle to enable or disable an intent without deleting it.
Consider “Transfer at End‑of‑Call” if the AI should gather information first and then transfer.
CAUTION:
“Transfer at End‑of‑Call” can create repetition if the partner re‑asks questions. Use it with trusted partners who accept AI‑gathered notes.
Test and Validate Your Intent
NOTE:
Test the shortest path first. The fastest valid path should always reach the correct action.
1) Run a Happy Path test (the most common scenario) and verify the final action.
2) Run Edge Case tests (for example, partial information, conflicting statements, or location unknown).
3) Confirm that only one condition is triggered per scenario.
4) Verify that the caller's questions are necessary and non‑redundant.
5) Review call notes to ensure the AI uses the caller’s exact words where appropriate.
6) Repeat tests for different times of day if you use time‑based conditions.
QA Checklist (Pre‑Publish)
[ ] Names and descriptions are clear and plain language.
[ ] Each condition is mutually exclusive and maps to one action.
[ ] Questions are minimal and decision‑driving (no extras).
[ ] Transfer numbers and canned messages are correct.
[ ] Time‑of‑day or other injections behave as expected.
[ ] Notes capture the caller’s own words where useful.
Publish and Monitor
1) Toggle the intent ON during a low‑risk window (for example, non‑peak hours).
2) Shadow the first calls to verify the question path and routing.
3) Capture feedback from call takers and supervisors and log it for iteration.
4) Make small, focused tweaks—then re‑test before continuing.
Migrate to ConfigV2 (Step‑by‑Step)
1) Export your legacy intents to the migration spreadsheet.
2) Place the child intent directly after their parent in the sheet.
3) Leave transfer/canned message cells blank for parent intents (children hold the actions).
4) Ensure each child is created in the same sheet as its parent.
5) Import the sheet; identical canned messages will be reused automatically.
6) Review imported intents in ANET and test each path before enabling.
TIP:
Old intents default to Single Path. Update them to Conditional Paths where branching is required.
Best Practices for High‑Quality Intent Design
- Use short, plain‑language names and full‑sentence descriptions.
- Make conditions mutually exclusive; prefer specific over broad.
- Group similar actions to reduce branching.
- Keep question lists minimal; every question should inform routing or notes.
- Document the “why” of each condition to speed future updates.
- Re‑test after any change; confirm the shortest valid path still works.
- Place notes and warnings before steps to reduce errors.
Glossary
Intent: The goal of the caller’s interaction.
Parent Intent: The main category for related scenarios.
Child Condition: A specific, routable situation under the parent intent.
Conditional Path: A branch that the AI follows when criteria are met.
Canned Message: A reusable text response sent to the caller.
Transfer at End‑of‑Call: The AI gathers info first, then transfers the caller.
Time‑of‑Day Injection: Logic that routes differently based on the current time.
FAQ
Question | Answer |
Can I edit child intent names in the UI? | No. They display exactly as written in the spreadsheet. |
Can I create a child intent for an existing parent from a different sheet? | No. Both must be in the same spreadsheet. |
What happens if two conditions are true at once? | Behavior may be unpredictable. Make conditions mutually exclusive. |
Can I keep using my old intents? | Yes. They’ll work as Single Path intents unless you update them to Conditional Paths. |
Support & Additional Resources: Support@prepared911.com