What patterns do ServiceNow AI Agents leverage ?

[AB51]

Hello, 

As i understand, ServiceNow AI Agents leverage the ReAct pattern to perform actions. They typically follow the T cycle of thought, action, and observation. Are there any other design patterns that the AI Agents in ServiceNow can leverage ? Is there a way to specify which pattern to use in an Ai Agent ?

[Tanushree Maiti]

Please refer this links, see if it helps you:

Hybrid AI Workflows: Choosing the Right Pattern fo... - ServiceNow Community

Advanced AI Agent Instructions Guide: ServiceNow E... - ServiceNow Community

[warren_chan]

In short, not yet. We have tested a few different methodologies, and the tradeoff seems to be around quality, speed/performance, and the amount of flexibility for customer agent prompts. This is an area that is consistently revisited, so I'd say check back on our features list often.

[M Naser]

Hi

Great question! To your first point... yes, ServiceNow AI Agents do leverage the ReAct pattern (Reason, Act, Observe) as their core reasoning loop, but that's not the only pattern at play.

Under the hood the agent framework also supports elements of Plan-and-Execute — where the agent can decompose a goal into a sequence of steps before acting which you can notice when building more complex multi tool agents in Agent Studio.

That said AFAIK there is no explicit configuration toggle to switch between reasoning patterns per agent. The pattern behavior is largely shaped by:

• Your system prompt / agent instructions so how you instruct the agent to approach a problem heavily influences whether it reasons step-by-step, plans ahead, or acts more reactively
• The underlying model  ServiceNow uses its own Now LLM and integrates with external models, and reasoning depth varies by model

So while you can't say "use ReWOO instead of ReAct" in a dropdown you technically can influence the reasoning pattern through prompt engineering and tool design.