Skip to main content

​
AI & Model Providers

Kantilever uses AI to power intelligent email triage, data extraction, and submission processing. This page explains how we work with AI providers and your options for customization.

​
How We Use AI

Kantilever leverages large language models (LLMs) to:
  • Triage submissions - Classify and route incoming emails
  • Extract data - Pull structured information from unstructured content
  • Generate responses - Draft replies based on desk rules

​
Zero Data Retention (ZDR)

We maintain Zero Data Retention agreements with all our AI model providers. This means:
  • Your data is never used to train models
  • Inputs and outputs are not stored by providers after processing
  • No human reviewers at providers see your content
  • Requests are processed and immediately discarded

​
Our ZDR Partners

ProviderZDR Status
Anthropic (Claude)Zero Data Retention
OpenAIZero Data Retention
Google (Gemini)Zero Data Retention

​
Bring Your Own Key (BYOK)

For organizations requiring additional control, Kantilever supports Bring Your Own Key configurations.

​
Benefits of BYOK

  • Direct billing - Pay your provider directly with no markup
  • Usage visibility - Monitor costs in your provider’s dashboard
  • Custom rate limits - Use your organization’s negotiated limits
  • Compliance - Satisfy internal requirements for direct vendor relationships

​
Supported Providers

  • Anthropic - Claude models
  • OpenAI - GPT models
  • Google - Gemini models

​
Setting Up BYOK

  1. Navigate to Settings > AI Configuration
  2. Select your preferred provider
  3. Enter your API key
  4. Test the connection
  5. Save your configuration
Your API keys are encrypted at rest and never logged or exposed. Keys are stored using AES-256 encryption with AWS KMS.

​
Data Security

​
In Transit

All API calls to model providers use TLS 1.3 encryption.

​
At Rest

Any cached or processed data is encrypted using AES-256 with AWS KMS-managed keys.

​
Data Minimization

We send only the minimum context needed for each AI operation. Full email bodies are processed, but no unnecessary metadata or attachments are included.

​
Enterprise Options

For organizations with specific compliance requirements, we offer:
  • Private model deployments - Dedicated model instances
  • Custom data residency - Processing in specific regions
  • Audit logging - Complete AI interaction logs
  • VPC connectivity - Private network connections to providers
Contact [email protected] for enterprise configurations.

​
Frequently Asked Questions

​
Does Kantilever train models on my data?

No. We never train models on customer data, and our ZDR agreements with providers ensure they don’t either.

​
Can I disable AI features?

Certain features require AI processing. Contact support to discuss options for manual-only workflows.

​
Which model does Kantilever use?

We use a combination of models optimized for different tasks. The specific models may change as we improve performance.

​
Is my data sent outside my region?

By default, API calls may route through provider data centers globally. Contact us for region-specific processing options.