Post-Event Recap

From AI Pilots to Production

Building a Hybrid Cloud Foundation with HPE GreenLake.

Executive summary

What leaders discussed

Enterprise AI is moving from isolated experimentation toward a more autonomous operating model built around agents, shared data, and coordinated workflows.

Enterprise AI is moving from isolated experimentation toward a more autonomous operating model built around agents, shared data, and coordinated workflows. The discussion emphasized that this transition is not primarily a technology challenge. It requires organizations to standardize tools, reduce shadow AI, redesign operating processes, and create the security and governance structures needed to move safely from sandbox environments into production.

A central tension is the need to balance experimentation with control. AI teams need enough freedom, infrastructure, and budget to explore new approaches, but businesses increasingly expect measurable returns within 12 to 18 months. Successful initiatives connect AI investments to faster delivery, lower operating costs, improved customer outcomes, or stronger risk management rather than focusing on models, tokens, or technical novelty.

The conversation also highlighted emerging operational challenges that will become more important as agents gain autonomy. Enterprises need visibility into which agents exist, what systems and data they can access, how much they cost, and whether their outputs remain reliable. At the same time, roles across engineering, product, operations, security, and leadership are becoming more fluid, increasing demand for professionals who can understand the business problem and guide AI-enabled solutions from end to end.

Moderator & panel

Perspectives from the room

The panel explored how enterprises can move from isolated AI experiments toward coordinated agents, governed access, sustainable economics, and end-to-end operating models.

Steven Fatigante
Moderator

Steven Fatigante

Global Hybrid Cloud & AI Strategist

HPE
Raj Badhwar
Panelist

Raj Badhwar

CIO

Systems Planning & Analysis
Aparna Chekuru
Panelist

Aparna Chekuru

Vice President of Software Engineering

Mastercard
Richard Jacik
Panelist

Richard Jacik

Chief Technology Officer

American Institutes for Research
Amarender Sardar
Panelist

Amarender Sardar

Director of AI

EchoStar Corporation
Sairam Tadigadapa
Panelist

Sairam Tadigadapa

Senior Vice President of Engineering

TransUnion

Key themes

What defines the agentic enterprise

The conversation consistently returned to five requirements for moving from isolated AI tools toward a durable enterprise operating model.

01 / OPERATING MODEL

AI Is Becoming an Operating Model

Organizations are moving beyond individual tools toward coordinated agents that participate directly in business and technology workflows.

02 / ECONOMICS

Experimentation Must Be Balanced with Economics

Teams need room to explore, but AI investments must demonstrate near-term value and avoid uncontrolled infrastructure or token costs.

03 / GOVERNANCE

Agent Governance Is an Emerging Enterprise Priority

Companies need centralized visibility into agents, models, identities, permissions, activity, and access to enterprise systems.

04 / DATA QUALITY

Data Quality Determines Agent Reliability

Unstructured, duplicated, outdated, or conflicting information limits the accuracy of AI systems and creates new governance challenges.

05 / LEADERSHIP

Roles and Leadership Expectations Are Changing

AI is blurring traditional functional boundaries and increasing the value of business judgment, architecture, creativity, and end-to-end problem solving.

Actionable takeaways

What enterprise leaders can do next

These actions translate the discussion into practical decisions for technology, data, security, operations, product, and business leaders.

Treat AI as an enterprise capability, not a collection of pilots

Define the operating model, ownership structure, standards, and shared platforms required to support AI across the organization.

Create controlled environments for experimentation

Give employees and researchers the freedom to test new approaches without exposing sensitive data or creating unmanaged production systems.

Tie additional AI spending to measurable value

Require teams to connect requests for more compute, storage, or tokens to lower costs, faster delivery, improved revenue, or reduced risk.

Establish an AI gateway and agent catalog

Route model, agent, and MCP activity through a centralized layer that provides discovery, monitoring, access control, cost visibility, and performance data.

Implement identity management for agents

Assign each agent an approved identity, authentication method, owner, permissions, and defined access-control boundaries.

Use standard patterns for lower-risk agents

Preapprove common use cases and escalate only agents that request unusual permissions, consume excessive resources, or perform high-risk actions.

Maintain human oversight for exceptions

Keep people involved when agents move outside established patterns, access sensitive systems, or make decisions with material business impact.

Improve unstructured data before connecting it to agents

Identify outdated, inconsistent, duplicated, or low-quality content before making enterprise knowledge available through AI tools.

Separate verified data from inferred outputs

Clearly distinguish authoritative source data from AI-generated conclusions so users know when additional validation is required.

Continuously evaluate and enrich enterprise knowledge

Use automated processes to identify inconsistencies, prioritize recent information, track provenance, and improve retrieval quality over time.

Be selective about runtime model usage

Using AI during development may create significant productivity gains, but applying token-based models to every production transaction can make costs unsustainable.

Redesign teams around end-to-end outcomes

Bring product, engineering, customer experience, security, data, and operations together to solve business problems rather than preserving narrow functional boundaries.

Develop more hands-on leaders

Senior leaders do not need to build agents themselves, but they must understand the technology, economics, risks, and architecture well enough to guide investment decisions.

Plan for a broader AI attack surface

Strengthen security as autonomous agents, easier access to advanced tools, and increasingly physical AI systems create new vulnerabilities.

Event moments

Inside the conversation

The evening combined executive networking, architectural perspectives, and a moderated panel focused on agents, governance, enterprise data, security, and sustainable AI economics.

Technology leaders connected over dinner and wine tasting before the moderated discussion.
Technology leaders connected over dinner and wine tasting before the moderated discussion.
HPE shared an architectural approach for governing infrastructure, agents, data, identity, and security.
HPE shared an architectural approach for governing infrastructure, agents, data, identity, and security.
Panelists discussed the operating, economic, and governance realities of moving enterprise AI into production.
Panelists discussed the operating, economic, and governance realities of moving enterprise AI into production.
Peer exchange continued around the tables as attendees compared priorities and operating models.
Peer exchange continued around the tables as attendees compared priorities and operating models.
The conversation brought together leaders across technology, data, product, security, and operations.
The conversation brought together leaders across technology, data, product, security, and operations.
2941 Restaurant provided a private setting for executive discussion, dinner, and wine tasting.
2941 Restaurant provided a private setting for executive discussion, dinner, and wine tasting.
View the full photo gallery ↗