In this comprehensive demo, we explored how Proxima, powered by Pebblo, enables secure, identity-aware access to enterprise data, along with powerful governance capabilities to enforce data compliance, access control, and policy enforcement at scale.
Role-Based Access with Two Personas: Bob and Alice
The demo introduces two user personas:
- Bob, a regional manager with admin privileges and broader data access
- Alice, an individual contributor with restricted access
This distinction serves to demonstrate how Proxima, underpinned by Pebblo, enforces role-specific data permissions across the system.
Security and Governance Dashboard
Logged in as Bob (admin), we explored the Security & Governance dashboard which provides:
- A high-level view of all ingested documents from sources like SharePoint, GitHub, and Confluence
- Insight into user activity, document sensitivity, and retrieval volumes
- Flags for anomalous users based on data access patterns
Administrators can drill into specific connectors and documents to view associated access groups, labels (e.g. financial data, PII), and identity permissions.
Configuring the Ecosystem: Connectors, Identity, and LLMs
Proxima's backend configuration includes:
- Connectors: Integration with enterprise tools across documentation, messaging, support, and project management systems with custom connector support
- Identity Provider Setup: Seamless integration with Google, Okta, or Azure AD for user-level identity enforcement
- LLM Configuration: Support for OpenAI, Azure OpenAI, AWS Bedrock, or even fine-tuned custom models
This flexibility allows enterprises to operate in their preferred AI and data stack.
Real-Time Prompt Governance in Action
Bob successfully prompts Proxima for confidential insights like company growth strategy. Proxima fetches accurate answers, warns about sensitive content, and traces the source document.
When Alice tries the same prompt, access is denied. However, she can still query permitted information (e.g., PTO policies), and Proxima intelligently compiles answers from multiple approved sources like Confluence and SharePoint.
Identity Mismatches and Guardrails
In scenarios where identity provisioning is flawed (a common enterprise issue), Proxima still prevents exposure:
- Alice inadvertently has access to a sensitive customer file
- She prompts for customer SSN and income data
- Proxima blocks the request using policy-level restrictions, despite the entitlement error
Bob, on the other hand, can access this data with a policy-defined warning, as his group is permitted to view PII.
Policy Framework for Precision Governance
The governance panel enables teams to:
- Define and enforce policies by group, topic, or label (e.g., block PII for all non-admin users)
- Set confidentiality levels via integrations like Microsoft Purview
- Comply with frameworks like HIPAA and GDPR via tag-based controls
- Restrict specific data sources per group for tighter oversight
Smart Policy Recommendations and Custom Topics
Pebblo also introduces advanced features:
- Policy Recommendations based on observed user behavior and group roles
- Custom Topics, where enterprises can define domain-specific subjects and train the system with example data to enforce new policies
AI-Enforced Policy
A cutting-edge alpha feature lets you delegate policy enforcement to an AI agent, which handles policy rules dynamically. This reduces manual overhead and scales governance in complex environments.
Final Thoughts
Proxima, built on Pebblo’s governance backbone, empowers enterprises to:
- Connect AI tools to business-critical data
- Control access through identity-aware guardrails
- Audit every retrieval and policy interaction
- Scale compliance confidently in an AI-driven workplace
Whether you're managing sensitive customer data or enforcing industry regulations, Proxima and Pebblo make it simple to govern AI access with precision, flexibility, and full visibility.
