The Router is our AI powered replacement of the switchboard service.
No longer do your customers have to suffer through listening to every single number entry when calling into your business, they can simply describe their reason for calling and me sent to the right place.
Configuring a Router Assistant
After creating a Natterbox AI component from the Routing Policies, simply select the Router App.
(Optional) Define the goal of the Router
We recommend you use the default to ensure the Router Assistant simply routes the call and doesn’t go beyond it’s scope - but this can be overridden to better fit your use case - see more on goal definition here
Set the initial message to the customer
Treat the Router Assistant like a receptionist
Give them a name
Provide the welcome
Clearly state you need the customer to describe their reason for calling
Example “Good morning, thank you for calling Natterbox. My name is Alice and I'm an AI, before I get you assistance can you please describe the reason why you're calling today”
Set the message to the customer when a route is identified
Set a friendly message for when the customer is off handed - for setting multiple different messages per route see here
(Optional) Set the domain specific knowledge to help with routing
The AI is great at routing a wide range of cases, but if you have common technical terms of Product names which aren’t widely known, then you can add them here to help give context
Example “Freedom, PCI Payments, Prompt Studio, AI Assist are all considered Products of Natterbox”
Define the Routes
Edit the default Route
Give it a new name to match the topic of the call (i.e. Sales, Service, Finance) - see more here on how best to define a route
Define how the AI should decide upon that Route
Use a handful of keywords to describe your route
Example - Sales Route “Everything to do with commercials, such as prices and payment plans”
Click ‘+ Add Route” and repeat the same steps for every required route
Enable and configure Human Escalation - see here for more on this
Hook the Routes to the respective destinations you want to forward calls to (i.e. Sales route is connected to a Sales Group Queue)
Hook the failed to a destination which commonly receives the most traffic or the team best to handle general customer requests
The AI will attempt a few times to determine what the customer need is, but after a few failed attempts it will forward the call down the failed hook, so be sure that this routes to your agents!
Defining a Route
AI Router works best when each route is clearly and thoroughly defined. The more detail you provide, the more accurately the system can match incoming queries to the right destination, reducing the chance of hallucinations or misrouted calls.
Each route consists of:
A short, descriptive name (e.g. Sales)
A detailed description of the types of queries that should be routed there
Make the route name as specific and relevant as possible. For the description, go beyond vague terms like “calls to sales” and instead describe specific intents. For example:
Less effective:
Routes calls to a sales team.
Better:
Routes calls where the caller expresses interest in buying, pricing, demos, or speaking to a sales rep.
Best:
Routes calls related to inquiries about purchasing new or used vehicles, requesting pricing or financing details, scheduling test drives, discussing trade-ins, seeking information on leasing or special offers, speaking with a salesperson, or expressing interest in custom orders. Route all other calls to 'General Service.'
Avoid Conflicts and Overlap
Avoid overlapping descriptions that might cause confusion about which route to choose. For example:
Finance Team: Handles calls about credit ratings, loan applications, payment plans, or financing approvals.
Pre-Sales Team: Handles calls from people interested in purchasing a vehicle, discussing financing options, or asking about their credit in relation to a purchase.
In this case, a query like “I need to talk to someone about my credit rating” could apply to both. To avoid conflicts, ensure route descriptions are mutually exclusive and clearly scoped.
Use the AI Assistant to help identify overlaps and suggest improvements to your route definitions.
Define different messages per Route
If you wish to have the Assistant read out different messages for different routes, leave the ‘Message to play to customer when route is identified’ empty.
Then insert an additional instruction onto the default goal, just amend prompt to match the Routes you’ve configured with the messages you wish the Assistant to read out :
[INSTRUCTION: If Sales been selected, then inform the customer ‘routing you to a Sales Agent now’. If Support been selected, then inform the customer ‘routing you to a Support Agent now’. ]
Defining the Goal Prompt
The Goal Prompt defines the assistant’s task. It guides how the AI responds and takes action using detailed Route Definitions and the admin-configured Domain Context.
Example Goal Prompt
You are routing a Customer's call to a department. Using the <options_list> as a reference, explain in 1 sentence how you can assist. Then proceed to understand the best OPTION for the Customer. Your task is NOT to solve the Customer's issue, but only to identify a OPTION.
Structuring the Prompt
Purpose
Clearly state the assistant’s primary goal. There’s no need to include persona details (like name or tone) here, as those will be handled by the Persona application you will have set up earlier in the workflow.
Example purpose:
You are routing a Customer's call to a department.
Contextual Integration
Use the Domain Context to give the assistant helpful background such as:
A summary of the conference
Scope of the knowledge base
Key details like dates, speakers, or session types
Constraints such as
Do not provide legal advice or pricing information
This improves the accuracy and relevance of the assistant's responses.
Example:
First, using information in <domain_context>, explain in no more than 2 sentences about the conference
Structure Your Prompt
Break down expected behavior into steps:
Step 1: Summarize the conference using <domain_context>
in no more than 2 sentences
Step 2: Ask the user what they want to know next
This structure ensures consistency in outputs.
Encourage Interaction
Wrap the prompt with an open-ended question to keep the conversation flowing. Conclude the prompt with an open-ended question or statement that encourages user interaction.
Example:
How can I assist in answering more questions about the conference?
Examples
Here are some other examples
Product Support
You are answering questions by looking up information about the AcmeVoice product in a structured knowledge base. First, using information in <domain_context>, explain in no more than 2 sentences what AcmeVoice is and what it’s used for. Then ask how you can assist the Customer with troubleshooting, setup, or feature-specific questions.
Internal HR
You are helping employees find answers about internal HR policies using a structured knowledge base. First, using information in <domain_context>, briefly summarize what types of HR topics are covered (e.g., time off, benefits, code of conduct). Then ask how you can assist the employee in finding the right information.
FAQ
Why is the Human Escalation Route triggering when the customer hasn’t asked for a human?
Human Escalation may be triggering if your routes and descriptions are not sufficiently well defined. For example if a user says Can I talk to Sales
and the Router simply has Sales
as the name of a route with little or no description, the Assistant may make the assumption that Sales is related to a Human and therefore the Human Escalation route is best.