AI Sales Agent

The AI Sales Agent is a built-in assistant in Salesnode that helps your team sell smarter and faster. It handles customer conversations, automates outreach and follow-ups, and assists sales reps with real-time insights and recommendations — so your team can focus on closing deals rather than managing tasks.

How to configure your agent

Salesnode lets you customize how the agent behaves and interacts with you in the workspace settings. Here, you can select the AI provider, choose the chat model, adjust creativity settings, set response length limits, and control how the agent uses tools with approval workflows. Proper configuration ensures the assistant works efficiently, safely, and cost-effectively for your needs.

Overview

Default Provider

The default provider setting determines which AI provider powers your assistant. You must add a valid API key to enable a provider for use.

Available providers include:

• OpenAI
• Anthropic
• Google AI
• Mistral AI
• GroqCloud

Each provider offers different AI models with unique pricing, speed, and quality characteristics. For example, some may be faster but less accurate, while others offer advanced reasoning at a higher cost. When you switch providers, you’ll need to select a new default model compatible with that provider.

Important points about providers

1. The provider you select controls which AI models and tools you can use.
2. Pricing and performance vary by provider, so choose one that fits your budget and needs.
3. Without a valid API key, the provider won’t be available for the assistant.

Default Chat Model

The default chat model is the AI engine that handles your conversations with the assistant. For example, a popular choice is gpt-4.

Models differ in reasoning ability, speed, and cost:

• Advanced models like gpt-4 provide better accuracy and handle complex tasks well but cost more and may respond slower.
• Lightweight models respond faster and cost less but might struggle with complicated queries.

Changing the chat model affects all future assistant interactions, so choose a model that balances quality and cost for your typical use cases.

Choosing the right model

1. For detailed analysis and complex problem-solving, pick a higher-end model.
2. For quick, simple replies or high-volume usage, lightweight models are more cost-effective.
3. Remember that model choice influences both your spending and user experience.

Web search

You can toggle on web search for certain models. Currently, the following models support the web search tool:

Temperature (0–1)

Temperature controls how creative or precise the assistant’s responses are.

• 0.0 means very deterministic and precise answers with little variation.
• 0.2–0.4 produces focused, professional responses ideal for most business tasks.
• 0.5–0.7 offers a balanced mix of creativity and accuracy (recommended default).
• 0.8–1.0 generates more creative, varied, and less predictable replies.

Higher temperature settings make the assistant more imaginative but less consistent. Lower settings increase reliability and accuracy.

When to adjust temperature

• Use lower temperatures for operational tasks like data updates or structured workflows to avoid mistakes.
• Use higher temperatures for brainstorming, marketing copy, or creative writing where variety is valuable.

Max Tokens

Max tokens set the maximum length of the assistant’s responses. Tokens roughly correspond to words or word parts.

• A higher max tokens value lets the assistant give longer, more detailed answers.
• A lower max tokens value keeps replies short and concise.
• Very high limits may increase response time and cost.
• Very low limits risk cutting off answers before they finish.

This setting caps the response length only—it does not limit how much conversation history the assistant remembers.

Approval Workflows (Tool Use)

When the assistant is allowed to perform actions like modifying records or creating campaigns, approval workflows add an extra layer of safety. With this enabled:

• The assistant will ask for your confirmation before executing any action.
• This prevents accidental changes or deletions in your workspace.

Tool approval workflows

1. Approving tool executions protects your data integrity by requiring explicit permission for write or delete operations.
2. They help you stay in control when the assistant uses powerful built-in tools.
3. They are especially important if the assistant has broad access to sensitive data or functions.

How to Configure Your AI Assistant

Follow these steps to set up your assistant’s configuration:

1. Open the AI Assistant sidebar by clicking the header toggle or pressing Cmd + K.
2. Click Settings at the bottom of the sidebar.
3. Navigate to the Assistant tab.
4. Select your Default Provider from the dropdown after adding the necessary API key.
5. Choose your Default Chat Model based on your needs.
6. Adjust the Temperature slider to balance creativity and precision.
7. Set the Max Tokens to limit response length appropriately.
8. Enable or disable Approval Workflows depending on whether you want the assistant to require confirmation before taking actions.