Set a reasoning tier for a Rovo agent

Agent reasoning refers to the depth of logic and iterative steps a Rovo agent takes to complete a task.

There are currently two reasoning tiers:

1. Quick answers: Speedy responses best for searching and simple questions (10 max. loop limit)

2. Think deeper: Multi-step reasoning suited for analysis and synthesis (60 max. loop limit)

Choose a suitable reasoning tier

While a simple task might only require a single pass for a fast response, complex problem-solving benefits from deeper reasoning cycles. This allows the agent to self-correct, use multiple tools, and verify its own logic before delivering a final response.

For example, a customer support team might setup a triage agent with a low reasoning to quickly categorize incoming work based on keywords.

Whereas a product team might build a backlog review agent with high reasoning that allows the agent to perform iterative loops to analyze work items, assess dependencies and tradeoffs, and provide a thorough view of upcoming work.

If you’re setting up autonomous or high-volume agents, carefully consider which reasoning tier is most appropriate for your task. These tasks can quickly consume credits.

When you select a specific model, the multipliers shown reflect the indicative level of AI credits consumed by that model compared to other available models. Multipliers are indicative and intended as comparative guidance rather than exact figures, so actual usage may vary.

More about Rovo usage allowance

Set a reasoning tier for your agent

When you first build an agent, we suggest having a clear task in task in mind, making it easier for you to define its reasoning from the get go. If things change, update the reasoning at any time.

To select your agent’s reasoning:

1. If you’re not already there, navigate to Studio from the app switcher ()

2. Select an agent you’ve created. Or, create a new Rovo agent

3. In the side navigation, select Details.

4. In the Reasoning section, assess which option suits your agent.

5. Select your preference: Quick answers or Think deeper.

6. 準備ができたら [公開] を選択します。

Understand recommended and available models

When you select a reasoning tier, you’ll find a Recommended model as a default selection. This is the model we’ve tested, and validated its compatibly with Rovo agents.

We’ll periodically test, reassess, and update which LLMs are available for you to choose from when selecting Think deeper reasoning. The recommended and available models may change over time to reflect these outcomes, and we’ll suggest alternative models if they present a better cost for their performance.

LLMs are also periodically deprecated by their providers. We’ll provide at least 30 days notice when this happens, before switching your agent to the nearest equivalent model.