FOUR LAYERS · ONE GOVERNANCE MODEL
GenAI is Venerable's enterprise Generative AI platform, designed to extract data, automate workflows, and generate insights using agent-based AI solutions. It is built to support production-grade AI usage in regulated environments by embedding governance, auditability, and human oversight directly into its core architecture.
GenAI treats AI as enterprise infrastructure — governed, auditable, and human-centered by default — not a collection of isolated tools bolted onto a business.
GenAI hosts multiple purpose-built AI agents, each aligned to a specific business problem, while inheriting common platform services such as security guardrails, observability, evaluation, policy-based access control, and Human-in-the-Loop (HITL) workflows.
Security, compliance, auditability, and traceability are foundational capabilities rather than after-thoughts. Every agent on the platform inherits these properties without additional implementation effort from the team building the agent.
All AI capabilities are delivered as agents that can reason, orchestrate tools, and interact with users or systems through standardized workflows. There is no ad-hoc LLM invocation — every AI capability is an agent with defined inputs, outputs, and governance controls.
AI augments decision-making rather than replacing it, with HITL review and override capabilities embedded across agents. High-risk or low-confidence outputs are automatically routed to human review before any downstream action is taken.
Every entry point into the platform, from human users to other systems.
- Web UI for user interaction
- REST APIs for system-to-system integration
- Python SDK for internal tools and notebooks
- Microsoft Teams App integration
- MCP & A2A integration
- Event, batch, and request-response execution modes
Home to GenAI's purpose-built agents — IntelliDoc, FundIntel, and DeepLens.
- Accepts domain-specific inputs
- Executes AI-driven reasoning and extraction
- Produces structured outputs
- Optionally routes results through HITL review
The common spine every agent inherits, so governance is never re-implemented per agent.
- Agent orchestration and execution
- Prompt and configuration management
- Guardrail enforcement
- Knowledge base and vector storage
- Evaluation, observability, and audit logging
A secure, scalable, fully serverless AWS architecture supporting enterprise-grade AI workloads with strong isolation, governance, and operational resilience.
Each GenAI environment is deployed into a separate AWS account, providing strong isolation across data, compute resources, IAM boundaries, and operational blast radius — keeping dev, test, and production workloads fully separated.
No long-running servers, event-driven and on-demand execution, automatic scaling, and built-in resiliency — powered by native AWS managed services.
GenAI uses AWS Bedrock as its primary managed LLM service, with Bedrock Guardrails applied as part of the execution pipeline so outputs adhere to enterprise and regulatory expectations.
Jenkins-based CI/CD pipelines build and deploy application code and provision cloud resources, using YAML-based templates for consistent, version-controlled, repeatable deployments across all environments.
The GenAI Platform Design Diagram (attached separately) is the authoritative reference for component boundaries, execution flow, service interactions, and control points for governance and HITL. No additional interpretation beyond the diagram is required.
A long-lived AI construct that orchestrates multiple steps, delegates work to tools, applies business logic and validation, and produces auditable outputs. Agents are independent in function but standardized in behavior through shared platform capabilities.
Deterministic, callable functions agents use to retrieve data, parse and chunk content, validate and reconcile, and persist or enrich results — executed under agent control and logged for traceability.
GenAI uses the Strands framework as its primary agentic AI framework. Strands Agents act as orchestrators for multi-step workflows, coordinate tool execution, maintain execution context across steps, and support both synchronous and event-driven execution — implementing reason–act–observe loops for complex enterprise workflows such as document processing, portfolio extraction, and evaluation pipelines.
Governance is embedded into every layer of the platform. The four pillars work together to ensure every AI output is controlled, observable, and auditable.
A growing agent library delivers immediate business value, alongside an agent builder model that lets teams create new agents which automatically inherit platform controls — enabling scalable AI adoption without fragmentation or duplicated governance logic.