Deliverance AI and HPE launch Agentic Knowledge Solution for regulated use-cases requiring enterprise-grade governance, security and data sovereignty

• Joint architecture allows organisations to deploy AI agent teams within private-cloud, on-premises and air-gapped environments
• Validation covers intelligent model routing, multi-agent guardrails, human review, quality evaluation, auditability and data-sovereign retrieval
• At a UK mobile network operator, governed agents answered a root-cause network investigation question in under five minutes during a live engineering session, against the operator's estimate of up to three hours by hand

Deliverance AI today announced a validated solution combining HPE Private Cloud AI with Deliverance Agentic OS to help organizations deploy AI-driven business workflows in private enterprise environments. The HPE Agentic Knowledge Solution with Deliverance AI is designed for regulated and business-critical use-cases that require enterprise-grade governance, security, and data sovereignty.

Deliverance Agentic OS has been deployed and validated on HPE Private Cloud AI, part of the HPE AI Factory with NVIDIA portfolio, through the HPE Unleash AI partner program. The joint offering gives regulated organisations a tested foundation for applying AI-powered agentic workflow automation to business-critical work involving sensitive data while operating within HPE's governed AI environment.

The architecture supports private cloud and on-premises deployment, with air-gapped operation available for environments with especially stringent requirements. Sensitive information can remain inside the customer's security boundary, while each workflow operates under defined permissions, guardrails, human oversight, cost management and audit controls.

From AI experiments to governed enterprise work

The validation demonstrated how Deliverance AI workflows operate on HPE Private Cloud AI while leveraging enterprise governance, model choice, operational controls, and secure data access. Key workflow capabilities validated included:

• Workflow-optimized model selection: Deliverance AI SmartRouter directs each task to an appropriate approved model based on complexity, cost, performance and governance requirements. Organisations can use NVIDIA NIM, self-hosted models and other approved models, using lower-cost models for routine work and higher-capability models where needed. This helps reduce inference costs, improve response times and avoid tying workflows to a single provider.
• Workflow automation with human oversight: Multi-agent orchestration allows specialised agents to work together within scoped knowledge collections and defined permissions. Guardrails cover inputs, outputs and communications between agents, including protection of personally identifiable information (PII), while administrators can stop active operations through a global kill switch. This allows organisations to automate more work while retaining control, demonstrate how agents have operated and provide evidence when responding to regulators, auditors and internal compliance teams.
• Automated quality assurance: Agent responses are independently assessed against customer-defined requirements and best-practice criteria. Work that falls below the required standard can be rerun or routed to a named person for review. This improves the consistency and repeatability of recurring business processes, reduces errors and rework, and keeps accountable people involved in decisions requiring human judgement.
• Secure enterprise knowledge access: Agents can search and use approved document collections without moving sensitive knowledge outside the customer-controlled environment. This allows organisations to apply AI to proprietary and regulated information while supporting data-residency, security and confidentiality requirements.

Every workflow follows the same governed process. Agents source information from approved systems, validate claims and outputs, consult named specialists when required, keep low-confidence work in draft, obtain human approval for defined decisions and record their actions for audit.

Four categories of enterprise workflow

The joint solution applies AI-powered workflow automation across four broad categories of enterprise work, helping organizations reduce manual effort while operating within HPE's governed AI environment.

• Respond: Complete RFPs, tenders, security questionnaires, supplier qualification forms and due-diligence requests using approved company information.
• Produce: Prepare recurring materials such as audit evidence packs, board papers, quarterly business reviews and regulatory filings, with statements linked to their supporting evidence.
• Review: Assess contracts, supplier documents, claims and underwriting information against agreed policies, clauses and evidence.
• Resolve: Investigate incidents using authorised logs and enterprise systems, develop a cited root-cause assessment and prepare internal or customer responses for human approval.

Give AI agents a job, not just a prompt

Many organizations have already proven that generative AI can answer questions, summarize information, and help employees work faster. Agentic AI raises the stakes. Agents can interact with systems, use tools, create business documents, investigate incidents, and take actions as part of multi-step workflows.

Greater autonomy creates greater responsibility. Enterprises need to know what information an agent used, which model handled a request, what actions it took, when a person approved a decision, and whether policies were followed.

Deliverance AI is the basis for the HPE Agentic Knowledge Solution. It provides an agentic operating system, designed to run and govern AI agents across regulated enterprise environments. The solution combines agents and workflows with a control plane for guardrails, personal data masking, human oversight, quality evaluations, auditability, and cost management.

Early results show what this means in practice. At a UK mobile network operator, governed agents answered a root-cause network investigation question in under five minutes during a live engineering session, against the operator's estimate of up to three hours by hand.

“Enterprises want AI agents to complete useful work, but they need the controls expected from any other production system. This validation shows how Deliverance Agentic OS workflows can run on HPE Private Cloud AI, providing that foundation. Organisations can put agents to work on sensitive and regulated processes while retaining control of their data, models, permissions and approvals. They can trace how an answer was produced, intervene when required and keep an audit record of the work.”

“Organizations don't need another AI pilot. They need trusted ways to apply AI to real business processes. HPE Private Cloud AI provides the governance, security, and operational foundation required for enterprise AI, while Deliverance AI helps automate workflows such as investigations, RFP responses, and compliance reviews. All while helping our customers manage their costs and sovereign requirements. Together, we're helping customers move AI into production with confidence.”

Built on optimized compute, networking, and storage, HPE Private Cloud AI comes pre-integrated with NVIDIA AI software and HPE AI Essentials – giving teams a shared control plane for models, tools, workflows, and agent sandboxes. Customers get sovereignty over sensitive data, predictable economics through token-based inference controls, and a validated entry point sized to their workload, with room to expand as adoption grows.

Notes to editors

High-resolution imagery is available to download here.

About Deliverance AI

Deliverance AI builds the Agentic Operating System for regulated enterprises, enabling organisations to deploy AI agents for sensitive, business-critical processes while personnel keep control of approvals. The platform combines ready-built agents, a governed runtime and forward-deployed engineers, while every model call is logged with the sources behind the answer. It runs in the cloud or on-premises, including air-gapped installations, so data stays within the customer's boundary. Deliverance AI is certified to ISO/IEC 27001:2022 and headquartered in London. For more information, visit www.deliverance.ai.

Media contact: media@deliverance.ai