Settings - How to Configure Your AI Agent's Model, Reasoning, Fallback, and Voice
Learn how to select the AI model, set conversation memory, configure reasoning effort, enable web search, set up the fallback behaviour, add custom fields, and configure voice settings.
The Settings tab gives you control over how your AI Agent thinks, responds, and handles situations it cannot resolve. It has two sub-tabs: Text (for chat settings) and Voice (for call settings).
How to Select the AI Model
Under the Model section, click the Model dropdown.
Select the model you want to power this agent.
GPT-4o Mini
Fast, lightweight responses for simple FAQs
GPT-5.4 (recommended)
Best intelligence for professional workflows — handles complex queries well
GPT-5.4 Mini
Balanced cost and quality for general use
GPT-5.4 Nano
Most cost-efficient option for high-volume simple interactions
GPT-5.4 is recommended for most use cases. If your agent handles mostly simple FAQ-type questions, GPT-4o Mini or GPT-5.4 Nano can reduce costs.
How to Set Context Messages
Under the Conversation section, locate the Context Messages setting.
Use the minus and plus buttons to adjust the number.
This controls how many previous messages the agent remembers during a conversation. More context = better continuity in long conversations, but higher cost per message. For most businesses, 20 is a good starting point.
How to Configure Agent Settings
Under the Agent section, you will find four settings:
Max Reasoning Steps
Maximum number of tool calls the agent can make per message. Increase if your agent needs to look up multiple pieces of information per reply
Include All Tools
When on (green), all enabled tools are passed to the agent with every message. When off, tools are passed selectively
Reasoning Effort
Controls how deeply the agent thinks before responding
Show reasoning summary in test mode
When on, the Test Chat shows a summary of the agent's reasoning steps — useful for debugging
How to Set the Reasoning Effort
Under the Agent section, click the Reasoning Effort dropdown.
Select the level that matches your needs.
Default (model decides)
The model automatically chooses the effort level
Varies
None (disable reasoning)
Agent responds immediately without extended thinking
Cheapest
Low
Minimal thinking — for simple queries
Low
Medium
Balanced thinking — for most use cases
Medium
High
Deep reasoning — for complex or nuanced questions
Highest
Use High only when necessary. For a simple FAQ-based agent, Low or Default is usually sufficient.
How to Enable Web Search
Under the Web Search section, locate the toggle.
Click it to turn it on (green).
When enabled, the agent can search the internet in real time to answer questions not found in its knowledge base. This slightly increases response time and cost.
How to Configure the Fallback (Default Intent Behavior)
The fallback defines what happens when the agent cannot confidently answer a customer's question. This is one of the most important settings to configure.
Under the Fallback section, click the Configure button.
Bot Studio opens with a visual flow editor.
A typical fallback flow looks like this:
Step 1 — Starting Point. Already configured. Set to trigger on "no intent match."
Step 2 — Add a Working Hours Condition block. This checks whether it is currently within your business hours.
The flow then branches into three paths:
During Working Hours path: 3. Add a Send Message block — for example: "Let me connect you with a team member who can help." 4. Add an Assign Agent block — select the team members and assignment method.
Outside Working Hours path: 5. Add a Send Message block — for example: "Our team is currently unavailable. We will get back to you as soon as possible."
On Holidays path: 6. Add a Send Message block with a holiday-specific message, if needed.
Save the journey.
Without a fallback configured, the agent will have no way to handle questions it cannot answer — the customer will be left without a response. Always set this up.
How to Add Custom Fields
Custom fields let the agent access real customer data — like their name, phone number, or tags — and use it to personalise responses.
Under the Custom Fields section, click the "Select custom fields..." dropdown.
Select the fields you want the agent to have access to.
For example, if you add the customer's name as a custom field, the agent can greet them by name instead of using a generic greeting.
Summary
In this article, you learned how to select the AI model, set conversation memory, configure reasoning effort and web search, set up the fallback behaviour for when the agent cannot answer, add custom fields for personalisation. In the next article, we will cover how to test your AI Agent before going live.
Last updated