Skip to content
The blog

WhatsApp and voice support in Customer Service Management: how it works

Atlassian Customer Service Management now takes WhatsApp messages and phone calls. See how the AI agent, routing, and setup work, and what to plan first.

contents
  1. What changed
  2. The use case: every channel ends in a work item
  3. How the AI agent handles intake
  4. How routing works
  5. What you need to set up WhatsApp
  6. What you need to set up voice
  7. Five decisions to make before you go live
  8. Where we help
  9. Sources

Atlassian's Customer Service Management now accepts WhatsApp messages and phone calls as support channels, next to the support website, email, and chat widget. WhatsApp support is released and native voice is in open beta, and both need a Premium or Enterprise plan of Service Collection.

This guide covers what each channel does, how the AI agent handles intake, how a conversation reaches a person, and what to decide before you switch either channel on. Every product fact comes from Atlassian's documentation and cloud roadmap, checked on November 16, 2026. The sources are listed at the end.

What changed

WhatsApp. You connect a WhatsApp number to a customer experience. Messages to that number arrive as conversations, in the same place as conversations from the chat widget and from calls. Atlassian's cloud roadmap lists WhatsApp support as released in Q4 2026.

Voice. Customer Service Management has a built-in voice channel. You give it a phone number, choose who answers, and calls route to a space. Atlassian labels it open beta, and its roadmap lists native voice support as coming soon for Q1 2027. If you want to keep your existing telephony, the Amazon Connect integration is still available.

The use case: every channel ends in a work item

Some customers would sooner message you on WhatsApp than fill in a form, and a phone call is still how many urgent problems arrive. Before this, teams on Atlassian often covered those channels with Marketplace apps or a separate contact center tool, so part of the conversation lived outside the support record.

Now each conversation has a matching work item in your space, whichever channel it came from. You can organize those work items in queues and put SLAs on them, the same way you do for other requests. The person who picks up a handoff reads what was already said, so the customer does not have to repeat it.

How the AI agent handles intake

Each channel has a support mode. There are three:

  • AI agent. The AI agent handles the whole conversation.
  • Support team. The conversation goes straight to a person.
  • AI agent and support team. The AI agent answers first and hands off to a person when it can't resolve the request.

WhatsApp uses the chat settings of its customer experience, so changing those settings changes how WhatsApp messages are handled. Voice sets the support mode per phone number.

On a call in the combined mode, the customer hears a disclaimer and your welcome message, then the AI agent answers. If it can't help, or the customer asks for a person, the call transfers to an available team member and the customer waits. On a handoff, the agent who accepts can see the full transcript of what the AI agent and the customer said.

In the AI agent only mode there is no person to hand off to. If the AI agent can't resolve the request, it runs the handoff you configured for it, such as a closing message or a follow-up request.

How routing works

Atlassian documents one assignment logic for chat, WhatsApp, and voice:

  • Conversations are offered to available agents in round robin order, one offer at a time.
  • Each agent has 120 seconds to accept or decline. An agent who does not respond is set to Unavailable.
  • A site admin sets the maximum number of active conversations an agent can hold. If every available agent is at that limit, the conversation waits for the next free agent.
  • If no agent accepts within 20 minutes, the customer gets your fallback message and the conversation ends. If no agents are available at all in support team mode, the customer gets the fallback message right away.

The documented logic is availability and capacity. Atlassian's pages do not describe routing by skill, language, or customer tier. You get that separation from structure instead: each customer experience routes to one space, and each WhatsApp number links to one customer experience. A separate number for a separate team is the way to send its conversations to its own space.

What you need to set up WhatsApp

  • A Premium or Enterprise plan with Service Collection.
  • An organization admin, with access to the Facebook account that administers your Meta Business Portfolio.
  • An existing Meta Business Portfolio and WhatsApp Business Account. Customer Service Management cannot create either one, and Meta has to verify your business information first.
  • Chat turned on and configured for the customer experience, because WhatsApp follows the chat settings.

The organization admin connects the WhatsApp Business Account in Atlassian Administration, under the site's Customer Service Management phone numbers, and adds the numbers there. A space admin then opens the customer experience settings, selects Channels, then Messaging apps, and adds the number. You can connect one WhatsApp Business Account per Jira site, with more than one number.

What you need to set up voice

  • A Premium or Enterprise plan with Service Collection. Voice is in open beta.
  • Voice turned on in the space's features. It is on by default for new spaces and off for existing ones.
  • A phone number. An organization admin can get one through Atlassian, local or toll-free, in the US, the UK, Australia, and Switzerland. You can also port an existing number through Atlassian Support or redirect calls from your carrier. A site can hold up to 10 numbers.
  • A call experience: the support mode, the space calls route to, the voice and welcome message, and whether the recording is kept on the work item.

Three limits to know. A call ends after four hours. Emergency calls are not supported. And during the open beta, agents take chats and calls from a desktop browser, not from mobile.

Five decisions to make before you go live

  1. Who answers each number. Decide the support mode per channel, and write down what the AI agent should hand off instead of trying to resolve.
  2. Coverage and the fallback message. If people are part of the mode, at least one agent has to be Available whenever customers can reach you. Write a fallback message that gives a next step, such as a form link or your hours.
  3. How you identify customers. WhatsApp and voice customers start as anonymous. Matching runs on phone numbers, and Atlassian recommends storing them in E.164 format, such as +14155552671. WhatsApp messages and calls are also exempt from the customer experience's customer access settings, so they are processed whatever organization the sender belongs to. Decide what the AI agent may share with someone who is not yet identified.
  4. Recording and transcripts. Calls are recorded to produce transcripts. A setting decides whether the audio is kept on the work item or discarded after the transcript. Confirm your consent and retention requirements before you choose.
  5. Usage and compliance. AI agent resolutions in Customer Service Management are a usage meter that Atlassian starts billing on December 3, 2026, so check that meter before you route a busy number to the AI agent. Our guide to usage-based pricing covers the settings. If you handle protected health information or need data residency, note that Atlassian's roadmap lists HIPAA compliance and data residency for Customer Service Management as coming soon, not released.

Where we help

We are an Atlassian Platinum Solution Partner. We set up service management on Atlassian: the channels, the AI agent's intake and handoff, routing and availability, and the queues and SLAs behind them. We also clean up customer and identity data first, so the people who call and message you match the records you already hold.

See how we run service management work, or talk to an architect.

Sources

The blog, weekly.

One email a week with what we published. No drip sequence, and you can leave in a click.

Get an agent readiness assessment

Fixed scope. You get a findings report across identity, platform, and governance, an ownership gap analysis, and a sequenced plan for closing it.