Bot setup for MAX: AI consultant and conversations with experts in one chat

The case study for MAX demonstrated a hybrid bot that combines two roles: answering standard questions as an AI assistant and routing complex inquiries to live experts. The user remains in a single chat, and the switch between automated response and human assistance happens without a separate channel and without revealing direct contacts.

Two Modes in One

The scenario is illustrated using a fictional medical service called "MedConsult". The logic is straightforward: if a client asks a standard question like booking, canceling an appointment, or finding instructions, the bot answers based on pre-loaded rules and FAQs. Essentially, it's a first-line support that handles the most repetitive requests and frees up specialists.

For services where expert time is expensive, such automation is particularly valuable: less manual routine work, faster initial response, and lower risk that a person will spend hours on identical answers. If the question cannot be resolved with a template response, the same bot switches to a different mode and begins working as a mediator between the client and the expert. The user writes a message, the bot requests the text and sends it to the appropriate specialist, and the response is returned to the same chat.

From the outside, this looks like a single continuous conversation without a separate "transfer to human" button and without a second window. This approach is especially convenient where companies consider it important not to give clients direct contact with a doctor, consultant, or support staff member.

How Forwarding Works

The key mechanism here is not complex integration, but using the standard reply function within MAX. When the bot forwards a message to an expert, it adds a service marker with the client chat identifier. The expert sees a regular incoming message and responds to it using the standard button, without a separate operator panel and without manual recipient selection. The case shows this with a simple example: a message for a doctor is supplemented with a line containing an address needed only for the routing system.

Tell the doctor that tests are ready

Client chat ID: 482910

This line serves as the address for return delivery. The expert simply clicks "reply" to such a message, writes their comment, and the bot reads the chat_id from the original text and sends the response to the appropriate user. The second part of the scheme is role differentiation by chat_id: if a message comes from an account in the expert list, the system expects a reply to a previously sent request; if not, the bot perceives the sender as a client. As a result, the entire structure is built on a single marker, reply functionality, and a couple of rules, without heavy server overhead.

What's Needed to Launch

The value of this case study is that it describes not an abstract idea, but a fully reproducible scheme. It can be adapted not only to a medical service, but also to legal consultations, educational platforms, B2B support, and any product with a stream of similar questions and an escalation point to a human.

To launch, you don't need a large stack: a bot, clear routing logic, and a carefully compiled knowledge base for the first line are sufficient.

• A messenger where the bot can programmatically see replies to specific messages
• A list of experts with their chat_ids for role separation
• A rule that adds a service chat identifier to the client's message
• A rule that extracts this identifier from the reply and forwards the response back
• An FAQ base or prompt for the first line so the bot can handle standard requests without human involvement

At the same time, the scheme doesn't automate everything. The expert still needs to be embedded in the process, scenarios for escalation need to be clearly described, and you need to ensure the bot doesn't try to answer where human responsibility is required. Testing on real conversations is needed, quality control of responses, and clear boundaries for where automation ends. But from a launch perspective, it's still a lightweight and inexpensive way to assemble hybrid support without a full-scale contact center, without a separate operator office, and without long development of new infrastructure.

What This Means

Such case studies show that the practical value of AI in messengers today lies not only in generating responses, but also in carefully routing communication. For business, this is a chance to maintain a single customer interface, reduce the load on experts, and do so without breaking the process with complex integrations.