Chatbot Integration for Business: Channel, Bot Type and Scope

Chatbot integration means connecting a chat bot to the channel where customers write to you (your website, WhatsApp, Instagram) and to the systems that actually do the work (CRM, orders, bookings, inventory). A bot that is not connected to any system can only recite frequently asked questions; an integrated bot answers "where is my order?" with that order's real status. So the real decision before setup is which channel the bot will run on and how deep it will go.

We cover the technical connection steps in a separate article. Here we focus on the decisions a business needs to make before starting a chatbot integration: scope, channel, bot type, cost drivers and the legal framework.

What levels of chatbot integration are there?

Not every business needs the same depth of integration. Thinking in three levels makes the decision easier.

• Level 1, the information bot: answers fixed questions such as opening hours, shipping times and return terms. It does not connect to any internal system and is the quickest level to set up.
• Level 2, the lead-capture bot: as well as answering questions, it collects the name, contact details and subject of a request, saves it to the CRM and assigns it to the right person. A customer who wants a quote leaves the request by chatting instead of filling out a form.
• Level 3, the transactional bot: checks order status, books appointments, starts returns and shows stock information. This needs a secure connection to your order, booking or ERP system.

The higher the level, the greater the benefit, but maintenance, testing needs and the risk of errors rise too. For most businesses the right start is to build level 2 well and then move to level 3 with the single most-requested transaction.

Which level suits which business?

• E-commerce: most incoming messages are usually about order status, shipping and returns. A level 3 integration that looks up orders is the step that pays off fastest here.
• Appointment-based businesses: asking about available times and booking are repetitive tasks; a bot connected to the booking system eases phone traffic. In healthcare, promotion rules can restrict what you may say: in Türkiye, for example, announcing fees and campaigns is restricted, so the bot's answers must be written to fit those rules. Check the equivalent rules in your own market.
• B2B manufacturers and distributors: sales cycles are long and requests are technical. The bot's job is not to answer everything, but to collect the request with the right details, log it in the CRM and pass it to the sales team.
• Service firms and agencies: qualifying incoming requests with screening questions (budget range, timing, scope) is the use that saves the team the most time.

Which channel should you start with?

Choose the channel based on where customers already look for you. A website chat window is the easiest start: setup is in your control and not bound by platform rules. WhatsApp and Instagram are channels customers are used to, but each has its own rules.

WhatsApp's business messaging policy says you may contact people only if they have given you their number and agreed to receive messages from you. The same policy requires businesses that use automation to offer a fast, clear escalation path: handover to a person in the chat, phone, email or a web support form. So a "bot only" setup on WhatsApp is not right from a policy point of view either.

Meta's Messenger and Instagram messaging policy requires you to tell users they are talking to an automated experience at the start of the conversation, after a long gap, or when the conversation moves from a person to a bot. A single greeting line such as "I'm the digital assistant for Company X" covers this.

Rule-based or AI chatbot?

Rule-based bots work with pre-written flows and buttons. What they will say is fixed and the risk of giving wrong information is low, but they do not understand questions outside the menu.

AI-powered bots understand freely typed questions and generate answers from your knowledge source. The user experience is more natural; on the other hand, the accuracy of answers depends on the quality of the knowledge source and on how well the bot's limits are drawn.

In practice the healthiest setup is usually a hybrid: transactions such as order lookup, bookings and returns run through rule-based, button-driven flows, while free-form questions are answered by AI. We explain how an AI bot is trained on a knowledge base in our article on AI chatbot integration for websites.

When should you not build a chatbot?

Not every business needs a bot at this stage. In these cases, doing something else first is the better move:

• If you get a few messages a day and the team answers them comfortably, adding a bot brings complexity, not speed.
• If the answers to frequent questions are not written down, the bot cannot find the right information. Write the answers first.
• If it is unclear who will update the bot after setup, a bot giving outdated information within a few months will hurt the customer experience.
• If the bot has no person to hand conversations to, unresolved conversations, especially outside working hours, turn into complaints.

What drives the cost of chatbot integration?

Giving a number would be misleading; two projects with the same name can differ widely. The main cost drivers are:

• The number of channels to connect and each channel's own approval process
• The integration level: information only, CRM records, or order and booking transactions
• Whether you use a ready-made platform subscription or custom development
• Model and usage fees if AI is used
• Meta's pricing on WhatsApp, which depends on message type
• Preparing the knowledge source, testing and improvement work in the first months

Microsoft's Copilot Studio guidance suggests the deflection rate as a way to measure a bot's return: the share of requests that a rep would normally handle but that are completed through self-service. Before deciding to build, counting which questions your team answers today, and how many times, is the most realistic way to estimate what that rate could be worth to you.

What about data protection?

A chatbot processes personal data such as names, phone numbers and order numbers. Basic steps are offering a short privacy notice link at the start of the chat, having the bot ask only for the information it needs, and deciding how long conversation logs are kept and who can access them. If AI is used, ask at the contract stage where the provider processes the data and whether it uses it for model training. If you serve customers in Türkiye, KVKK (the Turkish data protection law) applies, and the Turkish Personal Data Protection Authority's guide on generative AI and personal data is a good starting read.

In short: a decision checklist before you start

1. Write the bot's main job in one sentence: giving information, capturing leads or completing transactions?
2. List the 20 questions asked most in the last three months and how often each was asked.
3. Choose your starting channel; if it is WhatsApp or Instagram, read that channel's rules first.
4. Decide whether you will use rule-based flows for transactions and AI for free-form questions.
5. Define how, and to whom, conversations will be handed over to a person.
6. Assign an owner for the bot and an update schedule.
7. Settle the privacy notice and log retention period.

Once you have decided, see our article on connecting a chatbot to your site the right way for the technical steps. If you would like to define the right scope for your business together, see our AI integration services.