For years, WhatsApp Business operated as a structured messaging channel: text, templates, multimedia, buttons, lists, forms. Voice was outside the perimeter — if a conversation needed to escalate to phone, you had to pull the customer out of the chat into a call center, a landline or a hastily set-up Google Meet.
In early 2025, Meta closed that gap with the Calling API for WhatsApp Business: companies can receive and place native voice calls inside WhatsApp, keeping the customer thread, logs and brand identity intact. For teams already operating on the WhatsApp Business Platform, this isn't a new channel — it's the consolidation of the one they already have.
This guide covers which use cases it unlocks, what to consider before enabling it and how it fits with the AI flows you likely already have in production.
What the Calling API actually is
The API exposes three main capabilities:
- Receive inbound calls from the call button on the business profile, now with programmatic routing instead of the default that rejected the call.
- Place outbound calls within an active session with the user, without forcing them to leave the chat.
- Webhooks for call lifecycle events: ringing, answered, busy, ended — to integrate with CRM, lead scoring and reporting.
Unlike a classic call center, calls travel over Meta's infrastructure, not PSTN. The user sees them inside the same conversation they already had with the brand.
The five scenarios where voice moves the needle
After the initial rollout, the patterns that are working in LATAM concentrate on five scenarios:
1. Recovering conversations that got stuck in chat
A well-trained AI resolves 80-90% of inquiries, but the remaining 10-20% are cases where the customer is confused, frustrated, or the problem is ambiguous. Detecting that signal (negative sentiment, repeating the same question, high time on a session) and offering "want me to call you in 30 seconds?" turns lost conversations into closes. Before, you had to ask for the phone, schedule, and lose someone along the way.
2. Closing high-ticket sales
Vehicles, insurance policies, elective surgeries, dental treatments, financial products — products where the final decision doesn't close on chat. The AI pre-qualifies and gathers data, the rep closes by voice within the same thread. The insured no longer drops at "I'll pass you to an advisor who'll call you at 555-…" because the call starts from the chat with a single tap.
3. Onboarding for complex products
SaaS platforms, financial services, insurance — where the first post-purchase contact defines retention. AI presents the product, identifies understanding gaps, and offers a call with an onboarding specialist the moment it detects confusion.
4. Collections recovery
A WhatsApp message has a 95% open rate but lets the customer decide when and whether to reply. A native call, offered after several ignored messages, raises the contact rate without making cold calls from unknown numbers.
5. Insurance claims handling and medical emergencies
Insurance FNOL (First Notice Of Loss) flows benefit from an immediate call when AI detects a serious incident from the photos sent. Same with clinics receiving messages from patients with urgent symptoms — AI escalates to a call with human triage without losing chat context.
What to consider before enabling
The Calling API solves a technical problem, but opens three operational considerations:
Real human capacity. An outbound call requires a human available immediately. If your team is in narrow timezones, the "we call you now" promise breaks on weekends. Solution: AI only offers a call when someone is available (configurable windows) and schedules in other cases.
Quality and script. The agent receiving the call must arrive with full chat context — if they have to ask everything again, the customer perceives worse service than before. CRM integration and handoff with automatic AI summary are non-negotiable.
Compliance. In regulated sectors (insurance, healthcare, finance), calls may require recording, explicit consent and log retention. The Calling API exposes webhooks but you handle recording in your own infrastructure — verify local requirements before production.
How it integrates with existing AI flows
If you already have a conversational AI platform on WhatsApp (like Keebai), the Calling API integrates as just another event in the flow:
- AI detects an escalation signal (sentiment, complexity, ticket value).
- Queries human availability in real time.
- If there's availability, offers the call with a "Call me now" button.
- Webhook fires the call from the right human agent (skills, language, vertical).
- After the call, the summary returns to the thread automatically to continue support if needed.
The point is that the chat ↔ call ↔ chat transition is invisible to the customer. All information persists in a single timeline.
What it isn't
The Calling API does not replace a robust call center if your volume is thousands of calls/hour. It doesn't have advanced IVR, complex skills-based routing, or native recording with analytics. For high volumes, integration with platforms like Genesys, NICE, Twilio Voice or Five9 remains more appropriate.
What it does well: consolidate the channel for businesses that already live on WhatsApp and don't want to force the customer onto an external phone.
Next steps
If your operation already has relevant WhatsApp volume and you still lose 10-20% of conversations because you can't shift to voice, the Calling API solves that specific gap. The precondition is having a team that can receive the calls with context and speed — without that, enabling it just exposes the problem instead of solving it.