WhatsApp and Phone in One Inbox: How to Set It Up
Bring WhatsApp, SMS, phone calls and marketplace questions into one inbox: channels, AI first replies, team handover, templates, KVKK and what to measure.
Putting WhatsApp and the phone in one inbox means that whichever channel a customer uses, the conversation lands in the same list, joins the same customer history and is handled under the same team rules. Fewer messages slip through, customers don’t have to repeat themselves on every channel, and managers see all channels in one report. Setup takes seven steps: choose the channels, connect them, define what the AI answers first, write assignment and handover rules, set up tags and views, settle the KVKK side and measure.
For a short definition, see omnichannel inbox. In this article we go through each step, the channel rules and the metrics worth tracking.
Why one inbox?
In many businesses, customer communication grows in a scattered way. WhatsApp lives on a phone at the front desk, SMS in a separate panel, calls on the phone system and marketplace questions in seller dashboards. Each channel works on its own, but customers don’t separate them: someone who wrote on WhatsApp yesterday calls today and assumes yesterday’s conversation is known.
| Situation | With separate panels | With one inbox |
|---|---|---|
| Customer wrote yesterday, calls today | The agent can’t see the chat and asks everything again | Earlier conversations are in the same customer history |
| A message arrives while its owner is on leave | Nobody notices | Assignment and views make it visible to someone else |
| Two people answered the same question | The customer gets conflicting answers | You can see who replied and who is looking at it |
| A manager wants a weekly summary | Numbers are gathered from each panel | One report per channel |
| A marketplace question is waiting | Invisible until someone opens the seller panel | In the same list as every other message |
An example: a dental clinic
In the morning, a patient asks on WhatsApp about implant treatment; the AI agent replies with general information from the knowledge base and asks which days suit them for a consultation. In the afternoon the patient calls the clinic. The voice agent answers and takes the appointment request; when the patient asks about insurance coverage, it hands the call to the front desk with a summary. The receptionist sees the morning’s chat and the call summary in the same customer history and doesn’t ask the same questions again. A day before the appointment a reminder goes out; when the patient replies “I can’t make it”, the conversation is tagged “appointment change” and assigned to the right teammate. Every step stays in one record. For clinic-specific flows and KVKK notes, see our article on the AI phone assistant for clinics.
Which channels come together?
| Channel | Typical use | What to watch |
|---|---|---|
| Phone | Urgent, complex or sensitive matters | Inform callers if you record; add the call transcript to the customer history |
| Quick questions, appointments, order tracking | Approved templates for business-initiated messages and outside the window | |
| SMS | Reminders, short updates | Prior consent for commercial content |
| Telegram | Support through a bot | Matching the person to a customer record |
| Messenger | Questions from your Facebook page | Messaging window rules similar to WhatsApp |
| API channel | Website chat, mobile app | Passing the customer identity from your own system |
| Trendyol and Hepsiburada questions | Product, shipping and return questions | Answers are public and follow marketplace rules |
You don’t need to connect everything at once; start with the two or three channels your customers actually use. When the phone and WhatsApp share a number, matching the customer is easy. On Telegram and Messenger, you may need to link the person to your customer record during the first conversation.
The phone is different from the rest: it is real-time, and the call itself happens in the voice agent or the agent desk. What moves into the inbox is the call’s trace, meaning its transcript and summary. Without that trace, a single inbox is incomplete.
AI gives the first reply, the team takes over
A large share of written questions repeat: opening hours, address, delivery times, return conditions, appointment availability. For these, an AI agent can give the first reply based on your knowledge base. Write the rules down first:
- Topics the agent answers: questions covered by the knowledge base that don’t require personal records.
- Topics it leaves to the team: an explicit request for a person, complaints, payment and refund disputes, actions that need a specific customer record, and failing twice on the same question.
- The moment of takeover: when a teammate steps in, the AI should go quiet in that conversation. Two voices replying in the same thread confuse the customer.
- A cautious start: on a new channel, routing the AI’s replies to the team for approval before they are sent is a safe way to get to know the agent in the first weeks.
The logic is the same on the phone: the voice agent answers the call and, when a person is needed, hands it to the team with a summary (warm transfer). We cover handover triggers and a sample summary in warm transfer from AI to human agents.
Marketplace questions need extra care. The answer is published publicly on the product page, follows the marketplace’s content rules and, on some marketplaces, can’t be edited after it is sent. For these questions, an “AI drafts, the team approves” setup is a sensible starting point.
Team workflow: assignment, tags, internal notes and saved views
Without structure, one inbox simply becomes one big crowd. These tools keep the work manageable:
- Assignment: every conversation should have an owner. An unowned conversation is the one everybody assumes someone else is handling.
- Tags: keep a short tag list by topic (returns, shipping, appointments, complaints) and priority. If people don’t use the same tags, reports lose their meaning.
- Internal notes and @mentions: leave context for a teammate in notes the customer never sees, and pull a specialist into the conversation by name.
- Saved views: save the filters you use often, such as “Open conversations assigned to me”, “Marketplace questions awaiting an answer” and “Open conversations tagged complaint”.
- Bulk actions: tag, assign or close similar messages together, for example after an announcement.
- Snooze: snooze a conversation whose answer will only be clear tomorrow, so it comes back to the top at the right time.
WhatsApp template messages
On the WhatsApp Business Platform, a 24-hour customer service window opens when a customer messages you, and within it you can reply with free-form text. If the business starts the conversation or the window has closed, you need a template message approved by Meta. Templates are prepared in advance and submitted for review; they can be approved or rejected. The template’s purpose determines its category: marketing, utility or authentication. Details such as a name, appointment time or order number are added as variables.
Practical tips: keep templates short and single-purpose, don’t slip marketing content into a utility template, and make sure you have the recipient’s prior consent for commercial messages; for Turkish consent rules, see İYS. For ready-to-use text, see our WhatsApp and SMS reply templates.
A note on KVKK
A single inbox gathers personal data from different channels in one place. That is good for order, but it also concentrates responsibility.
- Permissions: not everyone needs to see every conversation. Limit roles and access to what each job requires.
- Retention: decide how long you keep messages, transcripts and recordings, and delete what has expired.
- Informing people: if calls are recorded, inform callers at the start; on written channels, link to your privacy notice in the first message.
- Transfers abroad: find out where messaging platforms and your software process data; transfers abroad follow separate rules under KVKK.
- Special categories of data: customers may send health information or documents by message. Decide separately who can see such data and how it is stored.
For the phone side, see our guide to KVKK for AI phone assistants and the entry on KVKK and call recording. This section is not legal advice; review your own situation with a lawyer.
Measurement: is the shared inbox working?
| Metric | How to calculate | What it shows |
|---|---|---|
| First response time | Time between the customer’s first message and the first reply, per channel | How quickly the team and the AI respond |
| Resolution time | Time from opening to closing a conversation | Workload and process issues |
| Conversations closed by AI | Conversations closed without a human reply ÷ all conversations | Knowledge base coverage |
| Handover rate and reasons | Conversations left to the team ÷ conversations the AI started | Missing information and rules |
| Waiting conversations | Number of open, unanswered conversations and the age of the oldest | Risk of messages slipping through |
| Channel mix | Each channel’s share of conversations | Staffing and channel planning |
| Repeat contact | Customers who write or call again about the same issue soon after ÷ closed conversations | First contact resolution |
Averages can hide a few conversations that waited a long time; look at the median first response time and the longest wait as well. Set targets against where you were before moving to one inbox.
Setup steps
- List the channels your customers use and who looks after each one today.
- Connect the two or three busiest channels first, and check that phone call transcripts are added to the customer history.
- Fill the knowledge base with frequent questions and keep changing details such as prices and opening hours up to date.
- Write down which topics the AI answers and which it leaves to the team.
- Prepare the tag list, assignment rules and saved views.
- Prepare WhatsApp templates and submit them for review.
- Settle permissions, retention periods and privacy notices.
- Send test messages and make test calls on every channel within the team.
- After the first week, read the metrics and update the knowledge base and rules.
One inbox in Speakline
Speakline messaging channels bring SMS, WhatsApp, Telegram, Messenger and your own API channel together with Trendyol and Hepsiburada customer questions in one inbox. On the phone side, a Turkish-speaking voice agent answers calls and, when needed, hands them to a human agent with a summary of the conversation (call center); once the call ends, its transcript is added to the customer’s conversation history.
On written channels, the AI agent replies using your knowledge base, or your team answers with saved replies. When a teammate replies in a conversation, the AI goes quiet there, and you can switch it back on if you want. On a new channel, you can route the AI’s replies to your team for approval before they are sent. Tags, internal notes, @mentions, saved views and bulk actions keep the team organized, and conversations and outcomes can be sent to your CRM through the API and webhooks. If you’d like to plan your channels with us, get in touch.