GreenLake Intelligence 2

Multi-agent workflow - Mesh of intelligent agents, acting 24X7 to reason and solve problems across your siloed IT estate.

ROLE
UX Architect
Cross-Functional Lead

TEAMS
HPE - CTO Office, HPE Labs (R&D)HPE GreenLake Cloud Platform
Storage, Network, Compute, OpsRamp,Professional Services, Digital Experience

Situation

Multiple Business Units (BUs) within HPE were developing Agentic AI solutions in silos, creating a high risk of a fragmented and inconsistent user experience (UX) across the GreenLake platform. GreenLake Intelligence was introduced as a unified "Agentic AI mesh" designed to act as a 24/7 operations team capable of reasoning and solving problems across a hybrid IT estate.

Task

Faster, proactive issue resolution of tickets, less handoffs and confidence in AI-assisted support.

Multi-Agent Synergy: Using distributed agents to handle repetitive tasks, analyze real-time data for better decision-making, and prevent single points of failure.
Edge-to-Cloud Automation: Leveraging deep infrastructure insights across the entire workflow to make GreenLake more adaptive and efficient.
Predictive Support: Shifting from reactive to proactive and preventive troubleshooting to resolve issues before they impact end users

Action

Aligned 7 product teams for common interaction strategy:

Design Thinking Workshop: Conducted a cross-BU workshop with 25 leaders across 7 BUs to define a shared, customer-centric UX strategy.
Validated MVP use case through cross-BU voting and leadership reviews
Common UX Archetypes: Defined universal models for human-AI interaction, including Conversational, Ambient, Planning, and Reasoning agents.
Collaborative Prototyping: Developed concept posters like "Squad AI" and "Know Me, Follow Me" to meet users where they are and ensure seamless context sharing between

Results

Platform Impact

Defined GreenLake Intelligence UX strategy and reference architecture. Established a shared, human-centered design model across HPE

Customer Impact

Faster, proactive issue resolution. Reduced context switching and cognitive load. Increased confidence in AI-assisted operations

Organizational Impact
Aligned 7 business units around a single UX vision. Positioned CX as a strategic growth lever, not a cost center. User‑Centered, AI‑Native Approach
Common design patterns

View Announcement

DISCOVERY PHASE - Customer Problem

Traditional troubleshooting is reactive, manual, time-consuming, and often requires "war rooms" involving multiple teams to correlate data across different tools. Fragmented interfaces lead to "swivel switching," where users must constantly change contexts between domain-specific tools (e.g., Storage, Networking, Compute), increasing customer frustration, cognitive load and manual effort

Agentic OS Platform

Led the experience strategy for Agentic OS Requirements presented to the HPE Board, using a proactive troubleshooting use case to align cross-BU technical workstreams with the GreenLake Intelligence vision.

Customer & Stakeholder Sessions

STORYBOARDING

Created vision storyboards and videos for the 3 scenarios to facilitate the discussion.

System design for multi-domain use cases

AI Design Patterns

Key Architecture Design Considerations

Human-in-the-loop
Human input is not an afterthought—it’s a structured component of context. Analysts, engineers, and domain experts can define high-level goals, constraints, and escalation protocols that are interpreted by AI agents during decision cycles.
Inter-agent communication
GreenLake Intelligence uses open industry protocols to standardize the language of collaboration between distributed agents, whether Hewlett Packard Enterprise or third party, such as ServiceNow, Salesforce, PagerDuty, and such. For example, a model detecting anomalous behavior in financial data can request enhanced resources from another agent responsible for dynamic workload orchestration—creating an emergent system of shared intelligence. Agentic AI with contextual autonomy
GreenLake Intelligence is designed to give agents the situational awareness they need to make intelligent decisions based on real-time infrastructure metrics and business logic. Interagent communications, based on MCP and other protocols, is critical to the operation and scalability of the system.

Infrastructure-aware models
With HPE and third-party infrastructure exposed to GreenLake Intelligence, agents can proactively request compute, storage, or network resources or adjust fidelity based on resource availability. This reduces cost, optimizes performance, and helps ensure SLA alignment.

Adaptable goal-driven architecture Agentic systems are driven by goals, but any system needs to be designed with flexibility to adapt to known goals today, and future goals that may not yet be known. Different goals may require the creation of different expert agents.

Why this matters
By leveraging a standards-based approach designed to scale as the industry evolves, GreenLake Intelligence helps ensure that models don’t operate in isolation but as part of a human-aligned, infrastructure-aware ecosystem. By leveraging GreenLake, organizations gain:
− Real-time collaboration between AI and humans
− Elastic, composable infrastructure aware of workload needs
− Scalable orchestration of distributed AI agents
− Built-in transparency and oversight to meet regulatory and ethical requirements

Expert agents: Software systems based on LLMs fed with enterprise infrastructure context (telemetry, logs, configurations). Agents are continuously active and autonomous.

Reasoning and agent orchestration: The system that orchestrates the chain-of-thought interaction across HPE and third-party agents. Agent action may be event-initiated or human-initiated. Shared context, memory, and intent across agents are enabled.
Adaptive copilots: Human-in-the-loop is enabled through a copilot interface system in GreenLake, allowing for interaction in human-friendly formats. Human operators define the level of autonomy granted to the agentic system. Copilot interactions may be optimized with the benefit of situational context for precision to the task at hand.
Governance and guardrails: Identity- and context-aware authentication of agent objectives and interaction

Feedback Received

It was very helpful to understand where different services/orgs are in their Agentic AI journey and what their end goals are. Also, get different perspectives on the troubleshooting use case in terms of challenges and opportunities. it did spur creative thinking on how to solve for seamless UX across different domains”

Rajrupa Nandy, GreenLake Platform PM

We learned what is happening across HPE in the area of agentic AI workflows and their use in different domains…We should certainly continue to let teams make progress in their domains, and start by connecting agents across domains so that not matter where the customer lands up, then can be directed to the right response. As a second phase, we could consider a GreenLake intelligence common entry point for anything agentic.”

Sachin Ganu. DT, Networking/Aruba

It was helpful to connect the dots within the company for people building real use cases. We might be using the term agentic quite loosely and need to understand the capability for what is possible and not possible. Orchestration is very important across the components to collect telemetry and resources. There was a good debate about organic growth and innovation vs what we need to balance.

- Puneet Sharma, Fellow, HPE Labs

OUTCOME & NEXT STEPS

Collaboration Model: Led to clarification of platform needs vs BU specific needs

Shared AI Patterns: Identified shared patterns for trust, progressive disclosure, confidence score. Set up a weekly PM+UX forum to create these patterns and publish these collaboratively.

GreenLake Architecture Clarified: Led to clarity in mental model - meet the users where they are instead of bringing them to a central place. This allowed for greater flexibility for the business units with opportunities to integrate seamlessly with other domain specific agents.

Announcements

Antonio Neri. “HPE’s new vision for hybrid IT is fueled by agentic intelligence at every layer of infrastructure, so enterprises can realize their boldest ambitions and achieve previously impossible levels of IT operations performance and efficiency.”

Discover Conference Keynote by Antonio Neri

Reflection

This was an extremely fun yet challenging project to bring together 7 siloed product teams and agree on a common mental model and shared patterns for the user experience strategy. Folks came into the workshop with a lot of hesitation about who's doing what and did not want to be told what to do by the platform team. Yet when we started doing the brainstorming around customer journey and share out of common customer challenges, the room shifted. Doing break-out groups with folks you haven't worked before allowed teams to step out of their comfort zones.
Eventually we saw several teams converge on the same mental model for troubleshooting - Keep the Users where they are. It allowed for flexible integrated approach and opportunity to align on common design patterns and archetypes in the future while leveraging from each other to build a common solution.
What I would do differently is to engage with the key decision-makers early in the initiative and work closely with them to align leadership goals, expected outcomes and visibility, reward to PM teams.

Page updated

Report abuse