AI Agent Not Responding
If your AI agent is not producing responses, timing out, or returning errors, here are the most common causes and fixes.
Timeout Errors
Long or complex prompts can cause timeouts, especially with larger models. Try breaking your request into smaller, more specific questions. If you are asking the agent to analyze a large dataset, narrow the scope — for example, ask about a specific month instead of an entire year.
Empty or Incomplete Responses
This usually indicates the model hit its output token limit. Try asking for a shorter response or specifying a format: "Give me a bullet-point summary in 5 points." You can also switch to a model with higher output limits from the agent settings.
"Model Unavailable" Error
AI models occasionally experience downtime or capacity issues. Kazi automatically falls back through a chain of models (Gemini > Claude > GPT), but if all models are affected, you will see this error. Wait a few minutes and try again. Check the Kazi status page for any ongoing incidents.
Agent Not Using Your Data
If the agent is not referencing your invoices, clients, or projects, make sure the relevant tools are enabled in the agent's settings. Go to AI Agents > [Your Agent] > Settings > Tools and verify that the data access tools (invoice lookup, client search, etc.) are toggled on.
Slow Responses
Response speed depends on the model selected and current demand. Gemini 3 Flash and GPT Mini are the fastest options. Claude Opus and GPT-5 produce higher-quality responses but take longer. If speed is critical, switch to a lighter model for routine queries and save premium models for complex tasks.