CRM lookup and update
Reads the customer record before speaking, and updates it during the call when the flow needs it, like a new address or a confirmed order.
In-call actions
What a voice agent does during the call
While the caller is still on the line, it looks up the record, books the slot, sends a link, starts a payment through your provider or hands over to a person. Once the call ends, after-call workflows take over.
Eight actions, during the call
Acts inside your systems, live
Each action below is a tool the agent can call mid-conversation, with the result told to the customer before the call ends. When the caller hangs up, the call closes with its outcome, reason, sentiment and next step: after-call workflows act on it and Score rates it.
Knowledge and policy lookup
Searches your knowledge base, price list or policies mid-call and answers from what it finds, not from memory.
Live warm handover
Hands the call to a person on your team with the conversation context attached, so the customer does not have to repeat themselves.
Links and forms by SMS or WhatsApp
Sends a link, a form or a request for a photo to the customer's phone without leaving the call, then reads the reply back.
Payments through your provider
Starts the payment in your payment provider's or bank's own flow, connected by API, and tells the customer the result once the provider confirms it.
Calendar booking
Checks real availability and books, moves or cancels an appointment directly in Google Calendar.
Virtual PBX routing
Transfers the call to the right queue, extension or team through DRING's virtual PBX or your own SIP trunk, following the routing rules you set.
Channel switch
Moves the conversation to WhatsApp, SMS or email mid-call when that is the better place to finish the task, with the same memory carried over.
Guardrails on actions
Every action, controlled
An action that touches a live system needs limits, not just capability. These are the same guardrails that apply across the platform's tools.
Read-only by default
Every integration starts able to look up data only. Nothing is written until a write action is switched on for that tool.
Per-tool sign-off for writes
Each write action, a CRM update, a payment, a booking, is enabled explicitly, tool by tool, never as one permission for everything.
Safe to retry
A repeated call to the same tool with the same details does not double-charge, double-book or duplicate a record.
Every tool call logged
What was called, with what parameters, and what it returned is recorded against the conversation for audit.
Transaction control for high-risk actions
Payments and other high-risk actions carry their own confirmation step and limits before the transaction completes.
Human handover any time
The customer, or the agent itself when it is unsure, can move the call to a person at any point.
Three calls, three actions
What it looks like on the line
Three short transcripts, each ending with a different action completed before the call closes.
The item arrived damaged, I want to return it.
Sorry to hear that. I am sending a WhatsApp link now so you can photograph the damage.
ActionWhatsApp link sent
StatusPhoto received
Got it, your return is approved. A label is on its way.
Can I just pay the rest now over the phone?
Sure. I am starting the payment with our payment provider. Their secure step is on your phone now.
PaymentProvider's secure step
ResultPayment confirmed
Payment received, your receipt is on its way by email.
I need to book a check-up next week.
Tuesday at 11 AM is open. Shall I book that for you?
CalendarSlot booked
ConfirmationSMS sent
Booked. You will get an SMS confirmation with the address now.
How actions are configured
From catalog to live tool
Adding an action to a live agent goes through the same test and release pipeline as any other change to how the agent talks or acts.
Choose the tool from the catalog
Pick the CRM, calendar, payment provider or messaging channel from the integration catalog.
Bind it to the agent
The tool is attached to the specific agent and the point in the conversation where it can be called.
Set permissions
Read-only, or a specific write action, is switched on explicitly with its scope and limits.
Test, then release
The action runs in simulated conversations first, then reaches the live line through the same staged release as a prompt change.
Monitor
Every call to the tool is logged and visible on the dashboard once the agent is live.
FAQ
In-call action questions
Can the agent take a payment without a person on the line?+
Yes. The payment runs through your payment provider or bank, connected by API, with the provider's own confirmation step and the limits you set. The agent tells the customer the result.
What happens if a write action fails mid-call?+
The agent tells the customer about the failure rather than guessing at a result, and the attempt is logged for the team to review.
Can actions be limited to certain hours or lines?+
Yes. Permissions are set per tool and per agent, so a write action can be scoped to specific lines, hours or agent versions.
Do all agents get all eight actions by default?+
No. Each agent is bound to the specific tools it needs, chosen from the catalog during setup.