DMLY’s AI Conversation powered by GPT (Open AI Responses) allows you to automatically generate intelligent replies to customer messages across WhatsApp, Instagram, Messenger, Telegram, and Live Chat.
It can:
Reply instantly to incoming messages
Retrieve answers from your uploaded business documents
Search the web (optional)
Update contact data
Trigger backend APIs
Hand over to a human agent when needed
Before using AI automation, make sure you have connected OpenAI:
👉 Go to Integrations → OpenAI
Configure:
API Key
Default model (optional)
Embeddings (for file search)
If you plan to use document-based answers, upload your knowledge base and enable embeddings under:
Integrations → OpenAI → Embedding
Go to Automations
Create a new rule or edit an existing one
Under Actions, select:
👉 AI Conversation powered by GPT – OpenAI Responses
Click to configure
The System Message defines how the AI behaves.
Example prompt:
You are a helpful and friendly assistant for our business. Respond clearly, professionally, and concisely. Provide accurate information based on our services. If unsure, ask clarifying questions. Offer further assistance at the end of each response.
You can use personalization variables like:
#contact_name
#workspace_name
Keep prompts:
Clear
Specific
Business-aligned
Short and structured
Define what happens if AI is unavailable.
Example:
“I will transfer the chat to a colleague to assist you. Please hold on.”
This ensures smooth fallback.
Under the Configuration tab:
Choose from available models (e.g., GPT-4.1, GPT-4, etc.).
Recommendation:
GPT-4.1 for high-quality business replies
Smaller models for cost efficiency
Controls response length.
Example:
300–500 for support
Lower for short replies
Defines how many recent messages AI uses for context.
Example:
5–10 for short conversations
15+ for complex support threads
Enable File Search to allow AI to retrieve answers from your uploaded documents.
You must:
Upload files
Create vector store
Add Vector Store ID
Best for:
FAQs
Policies
Pricing sheets
Technical documentation
Enable Web Search if you want AI to fetch real-time information.
Use carefully to manage token cost.
You can allow AI to:
Get contact information
Update phone number
Update contact stage
Add or remove tags
Update notes
Update custom fields
This allows AI to automatically manage CRM data during conversations.
Example:
If user says:
“I want a demo.”
AI can:
Add tag: Demo Request
Update stage: Qualified Lead
Assign to Sales
Under the Functions tab, you can define backend APIs for AI to call.
Provide:
Function name
Server URL
HTTP method
Headers
JSON function definition
Use cases:
Check order status
Fetch pricing
Book appointments
Retrieve availability
Create tickets
This enables advanced automation beyond simple replies.
Use the Playground tab to:
Test prompts
Simulate conversations
Validate tool usage
Confirm function calls
Always test before going live.
You can add a delay before AI responds to simulate a more human-like reply time.
Example:
2–5 seconds for natural feel
You can combine:
AI Conversation
Smart Follow-Up
Human Transfer
Example flow:
User sends message
AI replies instantly
If no reply after X minutes → Smart Follow-Up
If AI confidence low → Transfer to Human
Enable Override Human Agent if you want AI to respond even when chat is assigned to a human.
Disable if:
You want manual control during active support sessions.
Trigger:
When message received
Actions:
AI Conversation
Add Tag based on intent
Update Stage
If confidence low → Transfer to Human
File Search: ON
Contact Update Tools: ON
Web Search: OFF
Human Transfer Rule: Enabled
Add Tags: ON
Update Stage: ON
File Search: Optional
Follow-Up: Enabled
File Search: ON
Web Search: Optional
CRM Tools: OFF
