Security & Governance
Autonomous AI without giving up control. Built for enterprises, priced for small business.
The "Oh Sh*t" Button
One click. All agents stop. Instantly.
Click "Start Demo" to see agents in action, then try the Emergency Stop.
Try it yourself: Start the demo and hit Emergency Stop at any time.
Bring Your Own Key (BYOK)
We provide the car. You provide the gas. Your API keys stay yours, giving you complete cost transparency and data control.
Your M365 Tenant
Email, SharePoint, CRM data stays here
Outermind
Orchestration layer only
Your LLM API Key
You pay your provider directly
Cost Control
See exactly what you spend. No surprise overages.
Data Privacy
Your API key = your data policies apply.
Transparency
Token usage visible in your provider dashboard.
The Safety Gateway
An LLM-powered security layer that analyzes all outbound communications before they're sent. Prevent sensitive data leaks, inappropriate responses, and policy violations.
PII Detection
Automatically scans for credit cards, SSNs, bank accounts, dates of birth, passport numbers, and driver's licenses before any message is sent.
LLM Content Analysis
AI-powered semantic analysis detects sensitive business information, inappropriate tone, and context-inappropriate disclosures.
Recipient Classification
Automatically identifies internal vs. external recipients and applies different risk thresholds accordingly.
Risk Scoring
Every outbound message receives a 0.0-1.0 risk score. High-risk messages are held for human review.
Choose Your Protection Level
Disabled
For initial testing and debugging agent behavior.
Dry Run
All messages allowed through, but analyzed and logged. Perfect for tuning thresholds.
Enforced
Full protection active. High-risk messages held for human review.
Enterprise-Grade Security
Human-in-the-Loop Approvals
AI agents escalate critical decisions to managers via email. One-click approve or deny directly from your inbox—no portal needed.
Configurable Autonomy
You set the rules. Let agents send routine responses automatically, or require approval for everything—configure thresholds per agent, per action, per recipient type.
Complete Audit Trail
Every agent action is logged with timestamp, context, and reasoning. Every escalation and human decision is traceable. Export to your SIEM for compliance.
Role-Based Access
Control which agents can access which tools. Sales agents don't need SQL access.
Automatic Escalation Failover
If the primary approver doesn't respond within your defined window, requests automatically escalate to a backup manager. Nothing falls through the cracks.
Data Isolation
Your data never leaves your Microsoft 365 tenant. We orchestrate, you control the data.
Secure OAuth Integrations
Third-party connections like QuickBooks Online use OAuth 2.0 tokens with read-only defaults, complete audit trails, and configurable rate limits.
Frequently Asked Questions
Is my data used to train AI models?
No. Outermind uses a BYOK model, meaning your data remains isolated in your tenant and is never used for model training.
What happens if an agent makes a mistake?
You control the autonomy level. Configure agents to require approval for sensitive actions or let them handle routine tasks automatically. The Safety Gateway catches risky content, the Emergency Stop lets you pause all agents instantly, and the Escalation Router ensures humans approve critical decisions before they happen.
How does the human approval workflow work?
When an AI agent encounters a decision that requires human judgment—like a large discount request or policy exception—it sends an email to the designated manager with full context and one-click approve/deny buttons. The manager responds directly from their email without logging into any portal. If they don't respond within your configured timeframe, the request automatically escalates to a backup approver.
How do I control costs?
You bring your own LLM API key (OpenAI, Claude, Gemini, Grok, or Azure AI) and pay your provider directly for token usage. Our platform fee is fixed monthly, so you know exactly what you're spending.
What authentication methods are supported?
Outermind uses Microsoft Entra ID (Azure AD) for authentication. Your existing security policies, conditional access rules, and multi-factor authentication requirements apply automatically.
How is financial data protected with QuickBooks integration?
QuickBooks Online access uses secure OAuth 2.0 tokens with read-only access by default. Optional contact updates require explicit approval workflows. Every query is logged for compliance, and configurable rate limits prevent excessive access. Your financial data is never stored—only queried in real-time.
Is Outermind SOC 2 compliant?
We are currently in the SOC 2 Type II certification process. Contact us for our current security attestations.