A Telegram self-help bot: from free trial to paid access

Turning a psychologist’s method into a bot

A psychologist commissioned a Telegram bot to guide conversations using his method. Alongside the dialogue, it needed to offer paid access and accept payments. The client also needed an admin dashboard to manage users and payments and review statistics. The product brought together three parts: conversation, payments and administration.

The client supplied the method. I independently designed the technical solution, selecting the language model and building the conversation controls for the self-help bot. A standard AI chat interface was not enough: a model can readily follow an unrelated topic, while this product needed to guide the conversation through a structured approach. I discuss that distinction in my article on choosing between ChatGPT and a custom AI integration.

How the dialogue controls work

I needed to translate the client’s method into predictable conversational behavior. I built three complementary mechanisms.

The first is prompting. The model’s instructions account for the method, the conversation context and the current task. They do more than define a conversational role: they guide how each response is formed.

The second tracks conversation state and adherence to the method, detecting deviations and steering the dialogue back. It also manages the priority of issues raised during the conversation. These controls address both user topic changes and errors by the model itself.

The third automatically selects the reasoning depth for the next response. The bot determines how much reasoning the response requires, so the client does not have to adjust that setting manually. I also added caching to reduce token costs.

From a free trial to paid access

Users begin with a short free conversation. The bot then offers paid access and accepts payment. The invitation to continue becomes part of the dialogue rather than a separate sales process. My article on payments in Telegram bots covers payment handling and the questions worth resolving before development.

Administration and results

I built an admin dashboard for the client’s day-to-day work. It provides user and payment management, along with statistics the client can review and analyze.

The bot has been running for over two years. The client is satisfied with the result: a conversational product with payment handling and the tools to manage it.

A separate project: delivering purchased courses

For a different client who sells online courses, I built a Telegram bot with payments and an admin dashboard. This project had a different purpose: letting users take courses they had already purchased directly in Telegram.

Once a user starts the bot, they gain access to their purchased course. The bot then sends each video or other learning material on a schedule, supporting progress through the course. The client no longer has to send lessons to each participant manually.

Through the admin dashboard, the client manages courses and assigns them to users who have started the bot. Scheduled delivery is one part of the course experience, not a replacement for the full learning flow.

I also focused on keeping the application lightweight: the bot runs quickly and takes up minimal server space.

Similar tools, different user journeys

Both projects use Telegram for user interaction, payments and product access. Each has an admin dashboard that lets the client manage the service independently. Their purposes remain distinct: the self-help bot offers a trial conversation followed by paid access, while the course bot supports users through learning they have already purchased.