White-label AI voice agents: an agency delivery checklist

What is a white-label AI voice agent?

A white-label AI voice agent is a phone assistant delivered under an agency or client brand. The agency configures the conversation, connects business systems and supports the deployment while the platform provides the underlying tooling. Evaluate branding separately from operational requirements: a logo does not establish reliable booking, client isolation or a safe handoff.

Document the branding scope before you sell

VoiceForge’s hosted agent share page supports workspace branding and a switch that hides platform promotional links. Review that page with the client rather than promising complete brand removal across every screen. The custom-domain field stores a setting, but custom-domain serving is not currently implemented. Use a verified hosted share URL in the handover.

Check white-label pricing and shared limits

Growth is the first white-label plan at $149/month: 500 included minutes, 10 agents, five total workspaces, 10 integration connections and 5 concurrent calls. Enterprise starts at $499/month with 2,000 included minutes, 30 agents, 15 total workspaces and 5 concurrent calls. Ask sales about future capacity. Free and Starter do not include white-label branding.

• The agency workspace counts: five total workspaces can mean one agency plus four clients
• Client workspaces share the organization’s plan; limits do not multiply per client
• Browser tests draw from minutes, so include acceptance testing in the usage budget
• Free offers 10 browser-test minutes each month, with no phone numbers

Write a client brief with acceptance criteria

Choose one initial workflow: inbound reception, appointment assistance or consent-based follow-up. Record the caller’s goal, approved knowledge, required details, connected systems and conditions for escalating to staff. Define the expected outcome before configuring the agent so the client can assess the same workflow at sign-off.

• For reception, test business questions and requests outside the approved knowledge
• For booking, test unavailable times and a failed calendar connection
• For handoff, test what happens when the destination does not answer

Create the workspace and verify client access

Create a client workspace under the agency and configure its agents, knowledge and integrations there. Invite the intended client users with access appropriate to their responsibilities. Check the workspace each user can open and which actions their role permits before sharing credentials or granting editing access.

Make the deployment reviewable

Generate an agent from the approved brief, then review its versioned Agent Spec JSON with the client. Check goals, call flow, tools, guardrails, compliance policy and handoff behavior. Record the version approved for launch and repeat relevant tests when its instructions, knowledge or integrations change.

Test the actual phone and integration setup

Browser tests help review the conversation, but phone routing needs a separate test on a paid plan. Record who owns the number and carrier account, assign the intended agent and call the real number. Verify the calendar or other integration uses the correct client account. Test staff transfer on the selected runtime, including what happens when the destination does not answer.

Keep compliance in the workflow

VoiceForge applies compliance checks before outbound calls. For any follow-up, verify the permitted purpose, consent, DNC/DND status, opt-out state, calling window and required disclosures. Record who reviews these requirements for the client and stop the launch if the proposed workflow is blocked. A branded page does not change the permitted use of the agent.

Run a client acceptance test

Prepare a small set of calls and record the expected outcome before running them. Include a straightforward request, a changed appointment preference, an interruption, an out-of-scope question and an integration failure. Have the client check factual accuracy and the handoff experience. Re-run failed scenarios after editing the agent and keep the approved version identifiable.

• A booking is successful only when the calendar confirms it
• A transfer needs a clear fallback if the person is unavailable
• An unanswered question should produce an honest limitation or escalation
• A failed tool call must not become a spoken claim of success

Quote delivery costs without assuming resale margins

Separate your agency’s service fee from the platform subscription and included usage. Budget for additional minutes where needed, phone numbers, applicable carrier charges, messaging, external integrations and support time. Allocate shared costs across clients using measured usage and workload. Use the current pricing page and provider bills; the subscription is not an unlimited calling package or a guarantee of profit.

Deliver a handoff pack and agree on ongoing review

Give the client a named support contact and a process for updating business information. Agree who reviews transcripts, failed tools, unanswered questions and usage after launch. Confirm bookings against the client’s calendar; attendance and revenue require evidence from the client’s own systems.

• Approved Agent Spec version and business knowledge owner
• Test scenarios, observed outcomes and any unresolved limitations
• Tested phone number, carrier ownership and routing responsibilities
• Verified hosted share URL and client access roles
• Staff transfer destination and agreed fallback when unavailable
• Plan limits, billing responsibilities and next review date