SOFA v2 — Secure Organizational Foundation AI, built on fully managed AWS services. Now available in AWS commercial Regions and AWS GovCloud (US).
Enterprise AI Foundation
AI is spreading across every organization, and so are the integration costs and security gaps that come with it. Thorben's Enterprise AI Foundation gives you one secure, federated platform to deploy, control, and manage AI across your organization, programs, and applications, turning AI sprawl into a strategic advantage.
AI Hub to connect siloed data and systems
Your data stays in your own AWS account. People and AI agents see only what they are allowed to see. Choose from leading AI models from Anthropic, Amazon, Meta, OpenAI, and others.
Organization-Specific AI
We build, implement, and train. Your personas (AI assistants for each role), your knowledge (secure search of your documents), your workflows (custom AI agents), and your systems (SAP, Microsoft 365, and business applications).
Secure and Compliant by Design
Zero Trust and AWS Well-Architected principles. In AWS GovCloud (US), SOFA v2 is built on services authorized at FedRAMP High and DoD IL-4/5, and is designed to accelerate your ATO.
Performance-Focused AI
When, where, and how users need it. Higher user productivity, better use of resources, capabilities you can redeploy, and AI that is easy to use — in chat, in applications, and through APIs.
Secure Organizational Foundation AI (SOFA) v2
SOFA v2 is the foundation on which your organization builds its own AI capability, using your data and your processes. It is built entirely on fully managed AWS services — Amazon Quick, Amazon Bedrock with Amazon Bedrock AgentCore, and Kiro — so AWS runs, patches, and scales the platform while your teams focus on outcomes. Thorben deploys SOFA v2 as code into your own AWS account, giving you a repeatable, secure base for AI assistants, secure knowledge search, and AI agents at the enterprise, program, or project level.
Capabilities
| Capability | What it does | Thorben implementation |
|---|---|---|
| AI Assistant & Agents | Staff ask questions, summarize, draft, and run multi-step tasks | Amazon Quick for business users; Kiro for technical users (no Quick license needed) |
| Model Choice | Use the right AI model for each task and switch as better models arrive | Amazon Bedrock |
| Secure Knowledge (RAG) | Answers from your own documents, with citations | Amazon Bedrock Knowledge Bases |
| Agent Engine | Runs agents securely in isolated sessions, with memory | AgentCore Runtime and Memory |
| MCP & API Integration | Connects agents to your systems through one governed gateway | AgentCore Gateway (managed MCP) |
| Identity & Guardrails | Agents act with the user's own permissions; tool use is deny-by-default | AgentCore Identity and Policy; Quick Spaces |
| Quality & Oversight | Every step traced; answers scored for correctness and safety | AgentCore Observability and Evaluations |
| AI-Assisted Development | AI administrators build and maintain agents through a reviewed pipeline | Kiro with CI/CD |
Key Features
Enterprise-Grade Standards
Consistent governance, security, and operations across every deployment.
Repeatable Deployment
Deployed as code, so each program gets the same proven, secure build.
Flexible Modular Access
Use AI through chat, custom agents, or APIs from your own applications.
Federated AI Orchestration
One hub coordinates agents and tools across teams and systems.
Security Guardrails
Need-to-know access, policy-checked tool calls, and a full audit trail.
Unified Platform Management
One place to manage assistants, knowledge, agents, and integrations.
Technical Architecture
SOFA v2 is a scalable architecture built from fully managed AWS services and deployed into your AWS account. Business users work in Amazon Quick, technical users work in Kiro, and applications call agents through APIs. Agents run in Amazon Bedrock AgentCore, draw on Amazon Bedrock models and your knowledge bases, and reach your systems only through the AgentCore Gateway, using each user's own permissions. AI administrators build new agents in Kiro and release them through an automated pipeline.
Deployment Options
AWS commercial Regions or AWS GovCloud (US), in your existing landing zone or a Thorben-built secure landing zone. An optional self-hosting module is available when you must run your own models.
Operations
Integrates with your AWS operations and observability for Day 2 management, or Thorben operates it for you.
Implementation Strategy
Thorben deploys SOFA v2 in your AWS accounts and can manage it on an ongoing basis. We start with discovery of your goals, pain points, IT landscape, data, integrations, and scope, then prove value with a pilot before you scale. One shared platform avoids duplicate data ingestion, shares AI across silos, and keeps costs transparent through AWS usage-based pricing in your own account.
AI Solutions We Offer to Run in SOFA v2 — Federal and Commercial
Software & Knowledge
- Software Development for SAP ERP
- Software Development for MS Power Apps
- Knowledge Management for SAP ERP
- Custom Processes
Business Operations
- Data Management
- Help Desk Assistant
- Acquisition Management
- Service Delivery
See SOFA v2 working on your own use case. Thorben offers demonstrations and pilots for federal and commercial organizations. To request one, use the Contact Us button below.
Feature availability varies by AWS Region. Amazon Quick, Amazon Bedrock, AgentCore, and Kiro are AWS services; Thorben is an AWS Partner.