Integrated architecture for secure, scalable AI
Viking combines its KONI platform, IBM watsonx and your existing systems into a coherent architecture – so data, models and workflows are managed as a single, governed capability rather than ad‑hoc experiments.
A reference architecture for regulated AI
Viking’s architecture defines how documents, data, AI models and applications interact across on‑premises and cloud environments, with governance built in at each layer.
Business & Workflow Layer
Digital workers, assistants and applications that support real processes such as onboarding, case management, reporting and service desk automation.
AI & Orchestration Layer (KONI + watsonx)
KONI provides drag‑and‑drop workflow design, local and API-based LLM access, and agent orchestration; IBM watsonx adds enterprise model management and agentic capabilities.
Data & Knowledge Layer
Document management, vector stores and governed data products provide the context AI needs, with local grounding for confidential content and watsonx.data intelligence for broader data governance.
Infrastructure & Governance Layer
Secure, monitored infrastructure (on‑prem, private cloud or hybrid) with watsonx.governance and Viking controls to manage identity, risk, compliance and observability.
Principles that make AI acceptable in regulated environments
The integrated architecture is built around security, transparency and interoperability, so you can add new AI use cases without re‑opening fundamental risk questions every time.
Sovereign and hybrid
The KONI platform can run on‑premises or in selected cloud regions, and can use local LLMs or external frontier models via secured APIs, depending on regulatory and data needs.
Data stays private
Confidential documents and knowledge bases are tokenised and stored locally, with vector databases and document stores on the platform, not on public services.
AI governance embedded
IBM watsonx.governance provides model inventories, policy enforcement and monitoring across agents and applications, rather than relying on manual spreadsheets and reviews.
Open and extensible
The stack is built on open source components (e.g. Ubuntu, ChromaDB, LLAMA.cpp) and open platforms, making it easier to integrate with existing systems and avoid lock‑in.
.webp)
Integrating line‑of‑business systems, documents and print
The architecture is designed to sit alongside your current systems, integrating via secure connectors to create end‑to‑end workflows rather than a separate AI silo.
.webp)
.webp)
Controls, monitoring and evidence built in
The integrated architecture makes it easier to answer the questions regulators and auditors ask: Where is the data? Which models are used? How are they governed? What guardrails are in place?
Data lineage and cataloguing
watsonx.data intelligence and Viking’s documentation capture where data comes from, where it flows, and how it is used in AI and analytics workflows.
.webp)
Policy enforcement
Controls can be applied centrally (e.g. disallowing certain data types as prompts, enforcing region boundaries) and pushed down into workflows and agents.
.webp)
Model and agent governance
watsonx.governance maintains a catalogue of models and agents, risk assessments, policies and performance metrics across the architecture.
Logging and observability
The KONI and platform layers log prompts, responses, tool calls, and system interactions, with the option to integrate into your existing monitoring and SIEM tools.
Architecture engagements with Viking
Current state assessment
Review existing infrastructure, data platforms, identity, security controls and regulatory obligations to understand constraints and opportunities.
Target architecture design
Adapt the reference architecture to your environment, selecting deployment patterns (on‑prem, cloud, hybrid), data integration approaches and governance tooling.
Use case mapping
Map priority AI and automation use cases onto the architecture, identifying required integrations, data sources and control points.
Implementation roadmap
Define a phased plan, starting with pilots and progressing to wider rollout, including responsibilities, timelines and risk management steps.
Ongoing evolution
Periodic reviews align the architecture with new regulations, use cases and technology options, keeping it sustainable over time.
Request an architecture and governance workshop
Bring together IT, security, compliance and business leaders for a focused session to map your current environment and outline an integrated architecture tailored to your organisation.