Phone: 0 (552) 380 25 25  |  Weekdays 10:00–18:00 · Technical support 24/7

🇹🇷 TR

Digital Bridge Blog

Artificial Intelligence

WhatsApp Business Chatbot: How to Answer Customer Messages 24/7 Without Losing Control

A WhatsApp business chatbot answers customer messages 24/7 from your own data. Business API rules, templates, Turkish data law and hand-over, step by step.

9 min read  · Digital Bridge Engineering Team
WhatsApp Business Chatbot: How to Answer Customer Messages 24/7 Without Losing Control

A WhatsApp business chatbot is software on the WhatsApp Business Platform (API) that answers customer messages from your own product, order and support data, and passes anything it cannot resolve to an agent with the full history. Built well, it handles order status and bookings around the clock; built badly, it traps customers in loops and creates data protection risk.

Your customers are already messaging you on WhatsApp

In Türkiye, customers contact a business through whichever channel is closest to hand, and that is usually WhatsApp. According to TurkStat's 2025 Household ICT Usage Survey, WhatsApp was the most used social media and messaging app among 16-74 year-olds, at 88.6%. Dealers send orders, customers chase deliveries, patients ask for appointments, and much of it lands on a sales rep's personal phone. Uses beyond appointments in healthcare are covered in AI in hospital operations.

The consequences are familiar. Messages sent after hours wait until morning, the same question is answered by hand dozens of times a day, and when the person holding the phone goes on leave, the conversation history goes with them. Because nothing reaches the CRM, nobody can see which customer asked what. Our customer service chatbot guide covers channel-independent bot design; this article focuses on the rules, technical set-up and Turkish regulation specific to WhatsApp.

The cost of doing nothing: the channel is growing without you

WhatsApp's weight in business messaging keeps rising. On Meta's Q1 2025 earnings call, the company said WhatsApp had passed 3 billion monthly active users and described business messaging as one of its main growth opportunities. Automation is spreading quickly too:

Meta said that in Q1 2026 more than 10 million conversations a week were being handled through Business AIs on WhatsApp and Messenger, up from 1 million at the start of the year. (Meta Q1 2026 earnings call)

In Türkiye, automated replies to customer messages are still far from the norm. In a survey of 781 e-commerce businesses published in the Turkish Ministry of Trade's E-Commerce Outlook Report 2025, 83.4% used AI to create product copy and images, yet only 24% used an AI chatbot for customer support. Customers are writing on a growing channel, yet at most businesses replies are still typed by hand.

Lost sales are not the only cost. Conversations scattered across personal phones cannot be audited or recovered when an employee leaves, so names, addresses, orders and sometimes health details sit outside the company's control.

What can a WhatsApp business chatbot do? Examples by sector

A WhatsApp business chatbot earns its keep on questions that repeat constantly and whose answers already sit in a system. Conversations that turn into judgement calls, negotiation or complaints should stay with people. How hotels apply the same split to guest messages is covered in AI in hotels and hospitality. The table below sums up typical uses by sector and the system the bot needs to reach:

SectorWhat the bot can handleSystem it must connect to
E-commerce and retailOrder status, parcel tracking, return steps, stock queriesE-commerce platform, courier system
Manufacturing and wholesaleDealer order status, price list requests, delivery datesERP, B2B portal
Healthcare and clinicsBooking, rescheduling, reminders, address and opening hoursAppointment system
Service and maintenanceLogging faults, technician visit times, job statusField service or helpdesk system

The rules to understand before building a WhatsApp business chatbot

A business WhatsApp bot is not installed on the WhatsApp app on a phone. Automation runs on the WhatsApp Business Platform (Cloud API), and Meta's rules shape the design directly:

TopicWhat it meansImpact on design
Business account and numberThe number is linked to your business account with Meta; using the same number in the WhatsApp Business app at the same time is limited and subject to conditionsDecide early which number the bot will own
Customer service windowAfter a customer writes, you can reply freely for a limited periodThe bot is built to respond quickly once the customer has written
Approved message templatesBusiness-initiated messages outside the window (delivery notices, reminders) use pre-approved templatesNotification wording is prepared and submitted in advance
Opt-inBusinesses message people who have agreed to hear from themRecord where and when consent was given

Türkiye adds two layers of law. The first is Law No. 6698 on the Protection of Personal Data (KVKK), Türkiye's GDPR-style data protection law: chats contain personal data, customers must receive a privacy notice, and you must know where the data is processed. If the platform or AI model sits abroad, that is a cross-border transfer; the rules in force since 1 June 2024 are covered in our guide to KVKK cross-border data transfer.

The second is Law No. 6563 on the Regulation of Electronic Commerce: promotional messages may count as commercial electronic messages, so prior consent and registration in İYS, the national message consent system, need to be assessed separately. Answering a customer's order query and sending a campaign are not the same thing.

Setting up a WhatsApp chatbot, step by step

  1. Classify existing conversations. Pull the 15-20 most frequent question types from the last three months of WhatsApp chats: order status, price, stock, returns, bookings, address. The bot starts with that list and hands the rest to people.
  2. Choose the bot's source of truth. Answers must come from your catalogue, FAQ documents and order and stock systems, not the open internet. The logic of grounding answers in documents is explained in our guide to RAG and enterprise LLMs.
  3. Integrate the systems. To answer "where is my order?", the bot must connect to ERP, e-commerce or courier systems. Those links are built through API integration, and order data is never shared without verifying who is asking.
  4. Write the hand-over rule. Complaints, disputed returns, a question misunderstood twice, or a customer typing "agent" all go to the team with the full history. Nobody should have to explain twice.
  5. Test Turkish, not textbook Turkish. Customers write with abbreviations, typos, regional phrasing and voice notes. The bot must treat a casual "where's my parcel" and a formal order status request as the same intent; our guide to Turkish natural language processing explains how. Voice notes also need a Turkish speech-to-text layer.
  6. Stop it making things up. The bot must never guess a price, stock level or delivery date. If the answer is not in its sources, it should say it is passing the chat to a colleague; the technical safeguards are in reducing AI hallucinations.
  7. Pilot, then measure. Start with one question family, such as order status, and a limited customer group. Track weekly the share of chats the bot closes on its own, the hand-over rate and first response time.

Three mistakes to avoid

  • The menu maze. "Press 1, press 2" flows wear out patience on WhatsApp. A bot that understands free text, backed by at most two quick-reply buttons, works far better.
  • Mixing campaigns into the bot channel. Unconsented bulk promotions break İYS rules and platform policy, and they can damage your number's quality rating.
  • Losing the conversation. If bot chats never reach the CRM, sales will not recognise the customer. Plan how records land on the customer card alongside the structure in our CRM and sales pipeline guide.

How we do this at Digital Bridge

We do not sell a boxed bot. We measure the problem first and build to fit:

  • Discovery and requirements analysis. We map your current WhatsApp traffic, the question types and which system holds each answer. You then receive a written proposal setting out scope, phases and cost.
  • Bot development. Through our chatbot and NLP service we build an assistant that grounds its answers in your catalogue, FAQs and order system. Depending on the need, it can use cloud language models or open-source models running on your own infrastructure.
  • Integration. We connect the bot to ERP, e-commerce, courier and booking systems through our API integration service, and conversations and requests are written to the customer card in your CRM software.
  • Dealer channel. If dealer orders arrive on WhatsApp, the lasting fix is often a B2B dealer portal, with the bot as the first touchpoint that points dealers there.
  • Data protection and regulation. Our data protection compliance consultancy prepares the privacy notice, consent records, retention periods and the cross-border transfer assessment. We decide which data may reach the AI model using the principles in AI and KVKK.
  • Pilot and roll-out. We pilot with one question family, tune hand-over rules on real conversations, then widen the scope.

If most customer requests still arrive by email, SmartMail, part of our Smart360 family, summarises incoming mail and classifies it by topic, so the WhatsApp bot and the inbox can be designed together.

Next step

You do not need a large project to begin. Roughly count last month's WhatsApp messages and write down the five questions that come up most. Then get in touch and we will work out together which of them a bot can take over and which system connections it needs.

For a broader introduction, read where to start with AI in business and our Artificial Intelligence hub.

Let us look at your case

Tell us about your process; after a needs analysis we send a written proposal with scope, phases and cost.

Request a Quote +90 552 380 25 25
Questions we hear most often

Frequently Asked Questions

What is the difference between the WhatsApp Business app and the WhatsApp Business API?

The WhatsApp Business app is a free app for small businesses, managed from a phone, that offers quick replies and greeting messages. The Business Platform (API) lets several agents share one number, connect a chatbot and integrate with systems such as CRM and ERP. A custom AI bot connected to your own systems needs the API.

Can a WhatsApp chatbot message customers first?

Within the service window after a customer writes, the bot can reply freely. Outside it, business-initiated messages must use templates pre-approved by Meta, and the person must have agreed to receive messages. For promotional content in Türkiye, commercial message consent under Law No. 6563 and registration in the İYS consent database must also be considered.

Can a WhatsApp chatbot comply with Turkish data protection law?

Yes, but not automatically. Customers need a privacy notice, you must define which data is processed, for what purpose and for how long, and access to conversations must be restricted. If the platform or AI model is hosted abroad, the legal basis for the transfer needs assessing. Special category data, such as health information, requires stricter safeguards.

How much does a WhatsApp business chatbot cost?

The main cost drivers are the number of question types the bot handles, the systems it connects to (ERP, e-commerce, courier, CRM), voice note support, the language model used and monthly message volume. The WhatsApp Business Platform's own messaging charges come on top. That is why a proposal setting out scope and phases after a requirements analysis is more reliable than a single list price.

Should you use an off-the-shelf chatbot tool or a custom WhatsApp bot?

If you have few question types and the answers never change, an off-the-shelf flow builder may be enough. If answers need live data from order, stock or booking systems, the bot must understand free-form Turkish, and conversations must land in the CRM, you need custom integration. Before deciding, pin down which systems the bot will connect to and where the data will be processed.

What does the bot do when it cannot answer?

A well-built bot does not guess. Under agreed rules, for example a complaint, a disputed return or a question misunderstood twice, it passes the conversation with its history to a live agent. Because the agent can read what the customer has already said, the customer never has to repeat the details.

Can it understand voice notes?

Yes. With a speech recognition layer that converts speech to text, the bot can transcribe a voice note and handle it like a typed message. Regional accents, background noise and product names all affect accuracy, so the feature should be tested on real customer messages, and any voice note it cannot understand should go to an agent.

Have a different question? Ask Us

Talk to an Engineer

Tell us what you need to solve. We'll come back with a written proposal.