Difficulties of the project
Texts could be sent and seen by all visitors. Human ingenuity is still so good and sometimes can be strange that even the smartest bot can be fooled. This means that forbidden topics could leak onto the screen.
We did double moderation. Bulk of moderation chat bot did it automatically. And then the second bot sent the phrases selected for publication to the client’s telegram chat for manual moderation
We made a lag of 5 minutes between messages from the user and we gently remind him about this.
So that one person does not capture the entire broadcast and does not overheat the second, manual stage of support, spammers also had to be filtered.
This is not news: the Telegram bot communicates by text. It means, that these texts should be good - we should not hide them over beautiful design or wow-mechanics. But there are a lot of messages: to explain rules, to warn about publication and to reject text
After the dialogue designer, the copywriter took over the texts for the bot. So they became not just a set of technical reactions but sounded humane and understandable.
How to communicate with users?