Conversation Context
Conversation context is the information from an ongoing interaction that helps an AI system understand what the user means, what has already happened, and what should happen next.
It can include previous messages, user details, workflow state, retrieved information, and actions already completed.
What Is Conversation Context?
Conversation context gives meaning to messages that would otherwise be ambiguous.
For example:
Customer:
"Do you have appointments tomorrow?"
Later:
"What about after 4?"
The second message only makes sense if the system remembers that the customer is asking about tomorrow's appointment availability.
That prior information is conversation context.
What Can Conversation Context Include?
Conversation context may include:
- Previous user messages
- Previous AI responses
- User identity
- Customer profile
- Current intent
- Collected information
- Search results
- Tool results
- Workflow state
- Previous routing decisions
- Handoff status
Not every item needs to remain active at all times.
Conversation Context vs. Context
Context is the broader set of information available to the model.
Conversation context refers specifically to information related to the ongoing interaction.
Context can also include:
- System instructions
- Documents
- External search results
- Tool definitions
Conversation context is one component of overall context.
Conversation Context vs. Context Window
Context Window describes how much information a model can consider at one time.
Conversation context describes the conversational information itself.
A long conversation may exceed the available context window.
The application may then need to:
- Summarize
- Drop irrelevant turns
- Store state separately
- Retrieve previous information later
Conversation Context vs. Agent Memory
Agent Memory can preserve information beyond the current conversation or task.
Conversation context usually focuses on the active interaction.
For example:
Conversation context: the user wants an appointment tomorrow.
Long-term memory: the user generally prefers afternoon appointments.
The distinction depends on system design.
Why Conversation Context Matters
Pronouns and References
Users often say:
- "that one"
- "it"
- "the second option"
- "same time"
Without context, these phrases are ambiguous.
Follow-Up Questions
Users expect the system to understand follow-ups without restating the full request.
Workflow Continuity
The system needs to know what information has already been collected.
Routing
The correct route may depend on the conversation history.
Handoff
A human should receive the relevant context when taking over.
Conversation Context and Routing
Conversation Routing depends on context.
A user may change intent gradually.
For example:
- Ask about a service.
- Compare options.
- Ask about price.
- Request a booking.
A routing system should recognize the progression rather than treating each message as unrelated.
Conversation Context and Tools
Tool results can become part of conversation context.
For example:
- The user asks for available times.
- The system checks availability.
- The tool returns 3 PM and 4:30 PM.
- The user says, "Book the later one."
The system must preserve the tool result to understand what "the later one" means.
Conversation Context and Human Handoff
A strong Human Handoff preserves the important parts of the conversation.
This can include:
- Why the customer contacted support
- What was already answered
- What information was collected
- What actions were attempted
- Why the handoff occurred
This reduces repetition.
Managing Long Conversations
Long conversations create context-management challenges.
The system may need to identify:
- Which messages still matter
- Which facts should be retained
- Which content can be summarized
- Which tool results remain relevant
- Which details are outdated
More context is not automatically better.
Conversation Context and Grounding
Business-specific answers may also require grounded information.
Conversation context provides the user-specific part of the situation.
Grounding provides the authoritative source information.
Both can be necessary.
For example:
- Conversation context: the customer booked yesterday.
- Grounding source: the cancellation policy.
The AI uses both to answer correctly.
Conversation Context in AskHandle
AskHandle workflows can preserve the context needed to manage an ongoing interaction across routing, AI Answer, search, questions, tool use, and Human Handoff.
Continuous routing can use the evolving conversation rather than relying only on the latest isolated message.
This helps different workflow components operate as part of one coherent conversation.