SpyderBot Docs
Search Sign in

SpyderBot Documentation

Creating Runs

A Prompt Explorer run is a single AI investigation.

Overview

Each run captures a snapshot of how selected AI models respond to one or more prompts under a specific investigation configuration.

Rather than simply generating AI responses, a run creates a structured investigation that can be analyzed, compared, and referenced throughout the platform.

Every investigation begins with a run.

Business Decision

Creating a run helps answer one strategic decision:

What exactly do we want to investigate?

The quality of an investigation depends on how well it is designed.

Before running an analysis, organizations should clearly define the business question they want to answer.

Before Creating a Run

Before configuring a Prompt Explorer run, identify the purpose of the investigation.

Common objectives include:

Recommendation Analysis

Understand why AI recommends one organization instead of another.

Competitive Investigation

Compare multiple organizations using strategically important prompts.

Prompt Validation

Evaluate how AI responds to newly created prompts.

AI Model Comparison

Compare responses across multiple supported AI models.

Optimization Validation

Measure how AI responses change after implementing optimization efforts.

A well-defined investigation objective produces more meaningful results.

Designing an Investigation

A Prompt Explorer run is built from several configuration components.

Prompts
↓
AI Models
↓
Agent Configuration
↓
Run
↓
Investigation

Each configuration choice influences the observations collected during the investigation.

The following sections explain these components in greater detail.

  • Prompt Sets
  • AI Models
  • Agent Count

Running the Investigation

Once the investigation has been configured, Prompt Explorer begins collecting observations.

During processing, SpyderBot coordinates AI interactions across the selected investigation settings.

When the run completes, Prompt Explorer generates an investigation workspace containing:

  • AI responses
  • Recommendation analysis
  • Citation analysis
  • Entity analysis
  • Cross-model comparison
  • Supporting evidence

The exact results depend on the investigation configuration.

Repeatability

AI systems are probabilistic.

The same prompt may produce different responses over time or across different AI models.

Prompt Explorer is designed to support repeatable investigations rather than assuming that a single response represents permanent AI behavior.

Organizations often perform multiple runs to:

  • Compare different AI models.
  • Validate optimization efforts.
  • Investigate unexpected AI behavior.
  • Monitor changes before creating a Prompt Observatory.

For additional information:

Best Practices

Begin with a Clear Question

Avoid investigating prompts without a defined objective.

Instead of asking:

"Let's see what AI says."

Ask:

"Why does AI recommend Competitor A instead of us?"

or

"How does GPT-5 compare with Gemini for this prompt?"

Clear questions produce more valuable investigations.

Keep Variables Consistent

When comparing multiple investigations, avoid changing multiple configuration variables simultaneously.

Consistent investigation settings make comparisons easier to interpret.

Investigate Before Monitoring

Prompt Explorer is designed for exploration.

Once an important prompt has been identified, move it into Prompt Observatory for continuous monitoring.

Save Strategically Important Runs

Runs investigating commercially important prompts often become valuable reference points for future comparisons.

Common Investigation Scenarios

Organizations frequently create Prompt Explorer runs to investigate:

  • Product recommendation prompts.
  • Competitor comparison prompts.
  • Purchase-oriented prompts.
  • Industry-specific questions.
  • Brand perception prompts.
  • Educational prompts.
  • Emerging prompt categories.

Each investigation helps explain a different aspect of AI behavior.

Related Concepts

To better understand Prompt Explorer investigations:

Related Pages

The next step in creating an investigation is configuring its components.

Continue with:

These pages explain how each configuration influences the quality and reliability of an investigation.