Prompt
A prompt is the input or instruction given to an AI model to guide what the model should understand, generate, or do.
Prompts can range from a simple user question to a detailed set of instructions that defines the task, format, constraints, examples, and source information.
What Is a Prompt?
A prompt tells an AI model what is expected.
Examples include:
"Summarize this document."
"Explain this policy in plain language."
"Classify this request as billing, support, or sales."
"Return the result as JSON."
A prompt can come from a user, an application, a workflow, or a system-level instruction.
Types of Prompts
User Prompt
A user prompt is the direct request made by the user.
For example:
"What is your cancellation policy?"
System Prompt
A System Prompt defines higher-level behavior and instructions for the model.
It can establish:
- Role
- Rules
- Tone
- Boundaries
- Output requirements
Developer or Application Prompt
An application can also supply instructions that shape how the model should behave inside a product.
Task Prompt
A task prompt describes a specific job.
For example:
"Extract the customer name, phone number, and preferred appointment date."
What Can a Prompt Include?
A prompt may contain:
- Instructions
- Context
- Examples
- Constraints
- Output format
- Source information
- Role definition
- Business rules
- Evaluation criteria
The best prompt depends on the task.
Prompt vs. Context
Context is the information available to the model.
A prompt is one part of that context.
For example:
Prompt: "Answer using the policy below."
Context: the policy itself, previous messages, tool results, and other information available to the model.
A strong prompt cannot compensate for missing source information.
Prompt vs. System Prompt
A user prompt expresses the user's request.
A system prompt defines higher-level instructions for how the model should behave.
For example:
User prompt: "Can I get a refund?"
System prompt: "Answer customer questions using only approved company information. Do not invent refund exceptions."
The system prompt shapes how the user request is handled.
Prompt vs. Tool
A prompt tells the model what to do.
A tool gives the model access to an external capability.
For example:
Prompt: "Check today's available appointment times."
Tool: a scheduling function that retrieves the actual availability.
The model cannot reliably produce current availability from prompting alone.
Why Prompts Matter
Prompts affect:
- Task interpretation
- Output style
- Format
- Level of detail
- Constraints
- Role behavior
- Use of source information
Clear prompts reduce ambiguity.
But prompt quality is only one part of AI system quality.
Prompt Limitations
A prompt cannot guarantee that a model will always behave perfectly.
Prompts should not be treated as substitutes for:
- Access controls
- AI Guardrails
- Tool permissions
- Workflow logic
- Grounding
- Validation
- Human review
If a restriction must be enforced, it should not depend only on natural-language instructions.
Prompts in AI Agents
An AI agent may use prompts to define:
- Role
- Goals
- Tool-use behavior
- Routing expectations
- Output format
- Escalation conditions
- Source-use rules
The prompt works together with orchestration, tools, context, and workflow state.
Prompt Chaining
Prompt chaining uses the output of one prompt as input to another step.
For example:
- Extract structured information.
- Classify the request.
- Generate the response.
Modern agent workflows may use this pattern, although some systems instead use broader orchestration and tool-based steps.
Prompts and Grounding
Prompts can tell a model how to use grounded information.
For example:
"Use only the supplied source and say when the answer is not available."
The actual reliability still depends on the quality of the source and whether the model follows the instruction correctly.
Prompts in AskHandle
AskHandle workflows can use instructions to define how AI Answer or other AI-driven nodes should behave.
The prompt can shape the response, while separate workflow components handle responsibilities such as:
- Document Search
- Data Search
- Routing
- Human Handoff
- Tool-enabled skills
This keeps prompt instructions from becoming the only control layer in the workflow.
Writing Better Prompts
Useful prompts are generally:
- Clear
- Specific
- Relevant to the task
- Explicit about constraints
- Explicit about output format when needed
- Consistent with the surrounding workflow
The best prompt is not necessarily the longest one.