Skip to content
ChatterfyHelp Center
HomeGeneralCRMAI ManagerTrackerPlatformsFAQUpdate log
Основы и возможности
AI Manager — overview
What AI Can Do in Chatterfy
How AI works in Chatterfy
Подключение AI
Внутреннее устройство
Промпты и переменные
  1. AI Manager
  2. Основы и возможности
  3. AI Manager — overview

AI Manager — overview

What AI Manager is in Chatterfy , what layers it has (Settings, Reply, Routings, Recognition, Models, Variables), and which articles to start with. AI Manager is an overview help section about artific

Updated June 21, 20265 min
What AI Manager is in Chatterfy, what layers it has (Settings, Reply, Routings, Recognition, Models, Variables), and which articles to start with.
AI Manager is an overview help section about artificial intelligence in the platform. The parameters that actually change the bot’s behavior are in the bot card and in the flow settings. The section brings together the platform’s key AI tools:
  • Automation & AI (CRM → Bot → Automation & AI)
  • AI Settings — the bot’s personality (CRM → Flow → AI Settings)
  • AI Reply steps — a smart reply via AI (CRM → Flow → Steps → AI Reply)
  • AI Routings — meaning-based routing (CRM → Flow → Steps → AI Routings)
This shows how these tools work together, the best order to study them in, and which sections to explore for finer tuning of the bot’s AI logic.

Навигация в продукте: часть настроек — на уровне бота (Automation & AI), часть — внутри конкретного flow (AI Settings, шаги AI Reply и AI Routings). Ниже в таблице и блоках «Шесть слоёв» указаны и статьи справки, и типичные пути в CRM.


What tasks AI handles

auto-replies to typical lead messages without an operator;
routing by the meaning of the reply (agreement, objection, country, intent, etc.);
recognition of voice messages, images, and documents with data extraction into the customer card;
converting media to text for subsequent flow steps;
pre-qualification and segmentation of users before handoff to an operator.




What’s inside (article tree)

Below is a logical map of the AI section. First come the basic materials for understanding AI capabilities and the general operating logic, followed by step-by-step setup, and then deeper analyses and optimization.
Article
Purpose
What AI Can Do in Chatterfy
AI capabilities: auto-replies, routing, working with media, data extraction
How AI works in Chatterfy
AI architecture and the message processing path
How to connect AI to a bot
Step-by-step AI connection and launching the first AI Reply
How to set up AI after connecting
Media, routing, variables, and basic setup
How to improve AI quality
Prompts, model selection, A/B tests, and improving replies
AI Settings — Bot Personality
Global Legend, Restrictions, Response Style
AI Reply — detailed breakdown of the step
Goal, Requirements, Legend, Reply Addition, Skip
AI Routings — Deep Dive
Meaning-based routing and how it differs from regular Conditions
AI Recognition — media processing
Media recognition and data recording in Custom Fields
AI Models — model selection
Selecting models for Reply, Routing, and Recognition
Bot Variables in AI Prompts
Using variables inside AI prompts
Prompt Patterns
Ready-made templates and approaches for different scenarios
What to do if AI makes mistakes
Error diagnostics and troubleshooting
Section principle: why → what → how. First, the purpose and the big picture ( What AI Can Do in Chatterfy, How AI works in Chatterfy), then basic setup (How to connect AI to a bot), then a deeper dive (How to set up AI after connecting), finally quality and cost (How to improve AI quality). In-depth breakdowns of layers — when fine-tuning is needed.


Six AI layers

Практичный порядок чтения: «Что умеет AI» → «Как устроен AI» → «Как подключить» → «Как настроить после подключения»; глубокие разборы открывайте, когда появится реальная задача под слой.
AI in Chatterfy consists not of a single tool, but of several independent layers, each responsible for its own task. You can use them separately or combine them depending on the scenario.
For example, sometimes AI Reply alone is enough — an intelligent response via AI (CRM → Flow → Steps → AI Reply) without AI Routings. And AI Recognition — media processing может работать вообще без генерации ответов — только для распознавания голосовых, документов или скриншотов и передачи данных оператору.can work without generating replies at all — only to recognize voice messages, documents, or screenshots and pass data to an operator.
Separating AI into individual layers helps you configure the bot’s behavior more precisely and control the number of AI calls.
Layer / section
What it is for and where to go deeper
AI Settings
Bot personality at the flow level: AI Settings — Bot Personality
AI Reply
Reply generation at a step: AI Reply — detailed breakdown of the step
AI Routings
Meaning-based branching: AI Routings — Deep Dive
AI Recognition
Media and data extraction: AI Recognition — media processing
Automation & AI
Bot models and technical toggles: Automation & AI
AI Models
Choosing models for tasks: AI Models — model selection
Bot Variables
Variables in prompts: Bot Variables in AI Prompts
Prompt Patterns
Phrase templates: Prompt Patterns
Diagnostics
If quality has dropped: What to do if AI makes mistakes and How to improve AI quality
Core AI components:
AI Settings — bot persona (CRM → Flow → AI Settings)
Controls the bot’s role, communication style, restrictions, and the overall tone of its replies. Configured at the flow level.
Learn more: AI Settings — Bot Personality.
AI Reply — smart reply via AI (CRM → Flow → Steps → AI Reply)
Generates replies to the user based on incoming messages. Usually, this layer consumes the most AI tokens.
Learn more: AI Reply — detailed breakdown of the step.
AI Routings — semantic routing (CRM → Flow → Steps → AI Routings)
Determines which flow branch to guide the user through based on the meaning of their message. It does not generate the reply text itself.
Learn more:AI Routings — Deep Dive.
Recognizes voice messages, images, documents, and other media, turning them into text or structured data.
Automation & AI (CRM → Bot → Automation & AI)
The bot’s technical AI section: selecting models, configuring AI features, enabling Recognition and related settings.
Learn more:AI Models — model selection.
Bot Variables in AI Prompts
Reusable variables for AI prompts and messages via {{variables.*}}.


How to reduce AI spending

AI features use the company balance because each AI call accesses the selected model. The more actively AI is used and the more messages are processed, the higher the spending.
The cost is primarily affected by:
the selected AI model
the amount of text that AI receives as context (for example, a long Global Legend or a large conversation history);
the number of AI steps per user message (for example, AI Reply separately, AI Routing separately, and AI Recognition separately).
To reduce spending, you usually:
use different models for different tasks (AI Models — model selection);
shorten the Global Legend and remove unnecessary instructions;
do not run AI unless necessary;
monitor actual spending in Billing — bot expenses.
Learn more about optimizing quality and costs in the article How to improve AI quality.


Where to start

If you want to understand what AI can do in Chatterfy in general, start with the article What AI Can Do in Chatterfy.
If you need to understand how AI works internally and how the different AI layers are connected, see How AI works in Chatterfy.
If you want to quickly set up a basic AI scenario, go to How to connect AI to a bot.
If AI is already connected but you need routing, media handling, or variables, see How to set up AI after connecting.
If AI responses are poor, unstable, or often wrong, review How to improve AI quality and What to do if AI makes mistakes.
If AI is consuming too much budget, start with the articles AI Models — model selection and Billing — bot expenses.
Need ready-made examples of prompts and wording? See Prompt Patterns.
If you encounter issues with routing or media recognition, see AI Routings — Deep Dive and AI Recognition — media processing.


FAQ

Regular bots work only with predefined triggers and templates. AI can take the conversation context into account, formulate responses based on meaning, follow instructions from AI Settings, and route the user through different branches via AI Routings.

The main AI settings and prompts in the product are usually configured in English. However, the bot itself can reply to users in any language — this is configured separately through style and instructions.

Using AI consumes the company balance, because each AI call sends a request to the model. The final cost depends on the selected models and the number of AI processing runs.

Yes. A common scenario is that AI handles typical messages and then transfers the conversation to an operator based on flow conditions.

No. AI handles routine questions and first-line communication well, but complex situations, complaints, and important negotiations usually remain with a human.

Not necessarily. A stronger model is often used only for the AI Reply step, while lighter and cheaper models are used for AI Routings and simple tasks. Learn more: AI Models — model selection.

The actual spend is shown in Billing — bot expenses. Additional optimization tips are in the article How to improve AI quality.


Related

How AI works in Chatterfy
How AI Works in Chatterfy Architecture overview: which AI layers exist, what each is responsible for, and how they work together inside a flow. The complete message path — from incoming text, a voice
What AI Can Do in Chatterfy
What AI can do in Chatterfy AI does not replace the flow: it works within the bot logic you have already configured and helps handle routine tasks, rather than “inventing a product from scratch”. If y
How to connect AI to a bot
How to connect AI to a bot Step-by-step guide: from choosing a model to the first working AI Reply step. Basic setup takes 30–60 minutes including testing in Demo space. This article takes you from “t
AI Reply — detailed breakdown of the step
Unlike a regular Send Message, AI Reply does not send prewritten text. AI generates a reply when a message is received from the user. AI Reply is used for: communicating with the user handling objecti
Automation & AI
The Bot Settings → Automation & AI section brings together AI technical settings at the level of the entire bot . Which models to use (separately for logic and for generating responses), how to proces
How to set up AI after connecting
How to configure AI after connecting it Extending the basic AI scenario. You don’t need to enable everything at once — connect only what solves a real task. When basic AI Reply is no longer enough Sym
How to improve AI quality
How to improve AI quality Iterative tuning: prompts, A/B, model selection, measurements. The “observe → hypothesis → adjustment → measurement” cycle is the only way to avoid turning tuning into random
What to do if AI makes mistakes
What to do if AI makes mistakes Typical symptoms of AI issues in Chatterfy and what to check first: hallucinations, the wrong language, errors , высокий расход, проблемы Recognition, Variables, задерж

Was this article helpful?

NextWhat AI Can Do in Chatterfy

On this page

  • What tasks AI handles
  • What’s inside (article tree)
  • Six AI layers
  • How to reduce AI spending
  • Where to start
  • FAQ
  • Related