Understanding AI Models for AI Agents in BoldDesk
When creating an AI Agent, administrators must select a model from the Model dropdown. The selected model serves as the AI Agent’s underlying intelligence and determines how the AI Agent interprets customer requests, processes information, and generates responses.
The Model field is mandatory. An AI Agent cannot be created without selecting a model.
The Model dropdown displays all AI models currently supported and available within your BoldDesk environment. Available models are organised by provider, making it easier to identify the source of each model and compare available options.
Different AI models are optimised for different scenarios and may vary in areas such as:
- Response speed
- Reasoning and problem-solving capabilities
- Instruction-following accuracy
- Knowledge retrieval and contextual understanding
- Multimodal capabilities, such as processing both text and images
Depending on your support requirements, you may choose a model that is better suited for handling simple customer enquiries, knowledge-based assistance, troubleshooting workflows, or more complex support interactions.
The available models displayed in the Model dropdown may change over time as new models are introduced, updated, or made available by supported AI providers.
The following sections describe the AI models available for use with AI Agents, grouped by provider, along with their capabilities and recommended use cases.
Amazon Bedrock / Anthropic Claude
Use these models when the AI Agent requires Claude-based reasoning, enterprise workflow support, or fast response generation depending on the selected model.
| Model | General Purpose |
|---|---|
| Claude Haiku 4.5 | A lightweight Claude model optimized for speed and efficiency, suitable for faster and more cost-efficient responses. |
| Claude Opus 4.8 | A stronger Claude model designed for deeper reasoning, coding, enterprise workflows, and agent-based tasks. |
| Claude Sonnet 5 | A high-performing Sonnet model built for coding, agents, and professional work at scale. |
Google Gemini
Use these models when the AI Agent requires Gemini-based reasoning, multimodal capabilities, or a balance between performance and cost efficiency.
| Model | General Purpose |
|---|---|
| Gemini 2.5 Flash | A Gemini model designed for good price-performance and low-latency tasks that require reasoning. |
| Gemini 2.5 Flash Image | A Gemini model optimized for image understanding, image generation, and image editing. |
| Gemini 2.5 Flash Lite | A fast and cost-efficient multimodal model suitable for high-volume, lightweight tasks. |
| Gemini 2.5 Pro | A more advanced Gemini model for complex reasoning, coding, and advanced tasks. |
| Gemini 3 Flash Preview | A newer Flash model focused on speed and strong reasoning performance. |
Azure OpenAI / OpenAI GPT
Use these models when the AI Agent requires GPT-based reasoning, professional support workflows, coding assistance, or different levels of response quality and cost efficiency.
| Model | General Purpose |
|---|---|
| GPT-5.2 | An earlier GPT-5 generation model. |
| GPT-5.4 | A frontier model for professional work, coding, reasoning, and agentic workflows. |
| GPT-5.4 Mini | A smaller GPT-5.4 variant intended for more cost-efficient usage compared with larger GPT-5.4 models. |
| GPT-5.4 Nano | A lighter GPT-5.4 variant intended for lightweight or cost-sensitive usage. |
| GPT-5.4 Pro | A higher-performance GPT-5.4 variant for complex tasks requiring stronger capability. |
The models available in the Model dropdown may vary depending on the providers and models enabled for your BoldDesk environment.
Best Practices
Follow these best practices when selecting a model for an AI Agent:
- Choose a model that matches the AI Agent’s purpose.
- Use stronger reasoning models for complex troubleshooting or advanced workflows.
- Use faster or lighter models for simple, high-volume, or repetitive support queries.
- Keep model selection consistent for AI Agents with similar responsibilities.
- Test the AI Agent before deploying it to customer-facing channels.
- Review generated responses to confirm that the selected model fits the expected tone, quality, and accuracy.
Troubleshooting
The Model dropdown is empty
Confirm that AI models are available in your BoldDesk environment. The Model dropdown displays only the models that are currently enabled and available for selection.
A specific model is not listed in the Model dropdown
If a model is not listed, it may not be enabled or available in your BoldDesk environment. The available models may vary depending on the configured providers, such as Amazon Bedrock, Google Gemini, or Azure OpenAI.
The AI Agent responses are too slow
If response speed is a concern, consider using a faster or lighter model that is suitable for simple or repetitive support queries. Test the AI Agent before deploying it to customer-facing channels.
The AI Agent responses do not meet the expected quality
Review the selected model and test the AI Agent with real support scenarios. If the responses do not match the expected tone, accuracy, or reasoning quality, select a model that better fits the AI Agent’s intended support purpose.
Frequently Asked Questions
-
What is the Model field used for when creating an AI Agent?
The Model field determines the AI model that powers the AI Agent. The selected model affects how the agent understands customer queries, reasons through support scenarios, and generates responses. -
Is selecting a model required when creating an AI Agent?
Yes. The Model field is mandatory. An AI Agent cannot be created unless a model is selected. -
Why do I see different models in the Model dropdown?
The models shown in the Model dropdown depend on the providers and models enabled in your BoldDesk environment. Available providers may include Amazon Bedrock, Google Gemini, and Azure OpenAI. -
Can different AI Agents use different models?
Yes. Different AI Agents can use different models based on their purpose. For example, one AI Agent can use a faster model for simple questions, while another can use a stronger reasoning model for complex troubleshooting. -
How should I choose the right model for an AI Agent?
Choose a model based on the AI Agent’s purpose, expected query complexity, response speed requirements, and desired response quality. Test the AI Agent before deploying it to customer-facing channels. -
Does the selected model affect AI Agent performance?
Yes. The selected model can affect response quality, reasoning capability, response speed, and how well the AI Agent handles simple or complex support queries.
