AI Agents Inconsistents Results

[Al_Andrea]

Hi.

Is somebody else having troubles with AI agents?

The problem I am facing is that AI agents stop working for no particular reason and without any changes made to them.

For example, it can happen that I run the AI agent with some input and the first time it runs ok, then I run the AI agent again with the same inputs and the second time it doesn't work correctly anymore or tools that have always received the right inputs stop receiving the inputs. So some time the agents work, some times not.

Another example is the agent giving the error "Sorry, there was a problem on my side trying to complete this request. Try asking again later." without doing anything with no explanation whatsoever of what has gone wrong (not even in the logs of the AI Agent studio playground).

This is so random that I believe something is wrong with the system, but I already repaired a bunch of now assist plugins and nothing changed.

Some advice on how to make these agents work?

[abirakundu23]

Hi @Al_Andrea,

Yes, it is unpredictable behavior in AI Agents without changes any code in the system as a common experience.
This issue facing others developers while using AI Agents.
As a workaround, you can break into smaller task, refine Prompt, apply error handling mechanism like use log to track the issue and monitor the agent behavior.

Please mark helpful and correct answer if it's worthy for you.

[rpriyadarshy]

As you know AGENTIC AI is non -deterministic so results may vary little bit but its not running its strange.

When its not running woth same input and Prompt - Check these backend tables to get more Insights.

Execution Plans [sn_aia_execution_plan]

Execution Tasks [sn_aia_execution_task]

sn_aia_tools_execution

Regards

RP