Datasets · 10 min
Connect an HVAC Service Customer Dataset to a Voice Agent
Use a dataset when your voice agent needs structured customer records during calls. This walkthrough creates an HVAC Service Customers list with customer name, phone, address, service history, and tentative service date fields, then gives a voice agent read and write access.
Walkthrough
Step by step
Create the HVAC Service Customers dataset
Open Datasets, create a dataset named HVAC Service Customers, and use the description to explain that this table stores customer service history and tentative scheduling notes.
Create one dataset for the service customer list the voice agent should use.
Keep one customer or service location per row so the voice agent can read and update the right record.
Set up the columns
Rename columns to Customer Name, Phone, Address, Last Service Date, Last Service Type, and Tentative Service Date.
Rename the columns to match the fields the agent should read and update.
Use plain operational names. The voice agent will use these names when reading and writing dataset values.
Choose Phone as the AI Search column
In the dataset toolbar, set AI Search column to Phone. This tells the voice agent which field to use when finding a customer record during a call.
Choose Phone before attaching this dataset to a voice agent.
Phone is usually the best search key for inbound calls because callers can provide it quickly and it is less ambiguous than a name.
Attach the dataset to the voice agent
Open the voice agent, find Dataset access, and click Edit. In Available datasets, check Read and Write for HVAC Service Customers, then choose Attach.
Available datasets
Attach the dataset with both Read and Write enabled.
Read lets the agent look up service history. Write lets the agent create or update the tentative service date in the same dataset.
Tell the agent how to use the dataset
In the agent behavior instructions, tell the assistant to ask for the customer phone number, search the HVAC Service Customers dataset by Phone, summarize prior service dates, and only write tentative service dates when the caller asks to schedule service.
Instructions
When a caller asks about prior HVAC service, ask for their phone number. Search the HVAC Service Customers dataset using Phone. Summarize Customer Name, Address, Last Service Date, and Last Service Type. If the caller requests service, write their requested date to Tentative Service Date. Say the date is tentative until staff confirms dispatch.
The instructions should name the dataset, search column, and exact field to update.
Use the word tentative. The dataset should capture the requested date while your team still controls final dispatch or confirmation.
Test lookup and tentative scheduling
Run a preview call. Ask the agent to look up a customer by phone number, then ask for a tentative service appointment. Review Analytics or CRM to confirm the dataset tool calls appear in the transcript.
Can you look up my last AC service? My phone is 480-555-0142.
I found your February 18 service at 12 Mesa Ave. Would you like me to request a tentative service date?
Yes, tentatively June 12.
The transcript should show one dataset lookup and one dataset update when the caller schedules.
A good test call proves both paths: the agent can read the existing service history and can write the tentative service date back to the customer row.
Before you finish
Launch checklist
- The HVAC Service Customers dataset exists.
- Columns include Customer Name, Phone, Address, Last Service Date, Last Service Type, and Tentative Service Date.
- Phone is selected as the AI Search column.
- The voice agent has Read and Write access to the dataset.
- Agent instructions explain when to look up service history.
- Agent instructions explain when to write a tentative service date.
- A test call shows dataset lookup and dataset update tool calls in the transcript.
Fix common issues
Troubleshooting
The dataset cannot be attached to the agent.
Open the dataset and choose Phone under AI Search column before attaching it to the agent.
The agent cannot find a customer.
Confirm the phone number in the Phone column is entered consistently, then test with the same digits the caller would provide.
The agent reads the customer row but does not write a tentative date.
Open Dataset access for the agent and make sure Write is checked. Also confirm the instructions tell the agent which column to update.
Keep learning