Debugging

Debugging helps you identify, analyze, and fix issues in your AI’s logic, responses, or goal paths.
It’s your behind-the-scenes view of what really happened inside a conversation — why your Agent responded a certain way, which block or rule it triggered, and how it processed user input.

Think of Debugging as the AI’s black box recorder — every interaction, every decision, and every trigger can be traced back to improve accuracy and reliability.

Why It Matters

When your Agent doesn’t behave as expected — missing a goal, giving a weak response, or repeating lines — Debugging helps you uncover exactly why.

It allows you to:

Trace the logic path the AI followed.

Verify which Brain source or Journey block was used.

Identify missing or incorrect context.

Catch broken or misfired conditions before users do.

With Debugging, you’re not guessing what went wrong — you’re seeing it clearly.

Where to Find It

You can open Debugging from:

Sessions → Reviewing

Click the 🤖 robot icon next to any AI message.

The Troubleshooting drawer will open, showing technical details about how that specific message was generated.

What You’ll See in Debugging Mode

Section

Description

Message Context

The system’s internal understanding of user intent and conversation state before the AI responded.

Knowledge Source

Which Brain document, collection, or Q&A the AI used to form its answer.

Journey Path

Which Journey block or condition was triggered (e.g. “Hook → Align → Conversion”).

Conditions Matched

A list of IF/THEN logic checks and their outcomes (matched or skipped).

Version

The specific Agent version that handled the message — helpful when testing iterations.

Token Usage

(Advanced) How many tokens or steps the AI used to generate the response.

Error Logs

Warnings about missing context, failed goal triggers, or invalid actions.

Common Debugging Scenarios

Problem

Possible Cause

Debugging Fix

AI gave irrelevant response

Wrong or outdated Brain source

Reassign Knowledge collection or update context.

Goal not triggered

Condition mismatch or missing path

Check Journey logic; ensure block transition is defined.

AI repeated messages

Recursive prompt or memory loop

Reset memory or add stop conditions.

No response generated

Invalid Action or API error

Check integration (e.g., webhook or calendar task).

AI tone inconsistent

Conflicting Persona overrides

Verify Persona per block and global tone settings.

How to Debug a Session

Open any active or completed session in the Sessions tab.

Identify the message that needs inspection.

Click the 🤖 icon beside it — this opens Troubleshooting Mode.

Review:

Context – what the AI knew before replying.

Source – where it pulled its answer from.

Logic – which Journey block or action triggered.

Conditions – which IF/THEN checks were matched.

Use this insight to adjust your Brain, Journey, or Persona.

Recreate the same conversation (using the Recreate Session button) to verify your fix.

Pro Debugging Techniques

Technique

Description

Benefit

Version Comparison

Compare v1 vs v2 session paths to test AI improvements.

See measurable progress in tone, clarity, and conversions.

Tag Failed Goals

Use session tags like “Goal Missed” or “Debug” for tracking.

Organize your optimization workflow.

Recreate Session

Replay the same user messages after edits.

Confirm that your changes solved the issue.

Cross-Link with Knowledge

Check if the AI missed relevant content in Brain.

Fill Knowledge gaps to avoid future errors.

Real-World Example

During a session, your AI was supposed to trigger the “Book Demo” goal after detecting intent.
Instead, it continued the conversation without completing it.

After clicking the 🤖 Debug icon, you find:

Condition “Intent = Demo Booking” didn’t match because the user said “schedule a quick call” (not “demo”).

You update the condition to include synonyms (“book,” “call,” “schedule”).

You recreate the same session — and this time, the goal triggers perfectly.

That’s how Debugging turns insight into improvement — instantly.

Best Practices

  • Debug sessions after every major Journey or Brain update to catch regressions early.
  • Always test multiple user variations of key phrases (“demo,” “meeting,” “talk”).
  • Keep Knowledge sources clean and well-tagged for accurate context retrieval.
  • Use Agent Version labels to organize iterative testing.
  • Link Debugging findings back to your Playbooks and Guardrails.

Why It’s Powerful

Debugging doesn’t just fix problems — it builds intelligence.
Each time you trace and correct your AI’s logic, you’re training it to think, respond, and sell smarter.

Your AI doesn’t learn from perfection — it learns from the fixes you make here.

Robot Head Icon

Guardrails

Journey Advanced

Was this article helpful?