The Power of Agentic AI: Creating Value Through Business Transformation Roundtable
December 2, 2025 | Las Vegas
AI is rapidly transforming how organizations operate, compete, and plan for the future. Against that backdrop, technology leaders from semiconductor manufacturing, enterprise software, consumer retail, consulting, government, and hospitality gathered in Las Vegas for a candid conversation about the realities and the next horizon of agentic AI. What emerged was a picture of an industry at an inflection point, increasingly confident in the promise of autonomy but clear-eyed about the architectural, ethical, and operational challenges ahead.
Participants (in alphabetical order, by last name):
- Mohamad Ali, Senior Vice President, IBM Consulting, IBM
- Shibani Ahuja, Senior Vice President, Enterprise IT Strategy, Salesforce
- Ningyu Chen, Chief Technology Officer, Starbucks
- Eric Du, Senior Vice President, Chief Architect, OpenText
- Christine Gambino, Chief Operations Officer, Omni
- Ariel Kelman, President & Chief Marketing Officer, Salesforce
- Matthew Kropp, Managing Director & Senior Partner, BCG & Chief Technology Officer, BCG X, Boston Consulting Group
- Pawitter Mangat, Vice President, Global Tapeout and Mask Operations, GlobalFoundries
- Manish Mehrotra, Chief Information Officer, Hyundai Motor North America
- Suzanne Noble, Deputy Chief Information Officer, Information Technology Department, Clark County
- Glenn Nethercutt, Chief Technology Officer, Genesys
- Grisha Pavlotsky, Chief Transformation Officer, Miro
- Art Poghosyan, CEO & Founder, Britive
- Vikas Rao, Chief Technology Officer, Dark Matter Technologies
- Dan Regalado, Chief Information Officer, Wynn Resorts
- Mahdi Sajjadpour, Director, Solutions Architecture, Strategic ISVs and Tech, AWS
- Jeffrey Sullivan, Chief Technology Officer, Consensus Cloud Solutions, Inc.
- Akhilesh Tripathi, CEO, Digitate
- Bill Vass, Chief Technology Officer, Booz Allen Hamilton
Bloomberg Moderator
- Ed Ludlow, Co-Anchor, Bloomberg TV
Roundtable Highlights
Defining Agentic AI. Participants began by clarifying a point that often shapes the public conversation. Generative AI and agentic AI are not the same. For much of the past year, the word “agent” was used to describe tools that simply produced content. True agentic systems act on behalf of users, make decisions, and take action without constant intervention. Several leaders noted that this distinction has only solidified in recent months. One participant described it as a shift toward recognizing autonomy, rather than generation, as the next frontier of value creation.
Data Readiness as the First Barrier to Scale. The discussion then moved to a foundational challenge. Agents can only perform as well as the data they have access to. A GlobalFoundries executive emphasized that if an agent cannot see upstream suppliers or understand production variables, it cannot act meaningfully inside that ecosystem. In regulated industries, the expectations are even higher. Responses must be auditable and consistent, and they must align with compliance requirements. Without reliable data foundations, several participants agreed that agentic systems stall before they begin.
One participant cited the familiar statistic that 95 percent of AI projects fail, arguing that the failures usually stem from pilots built outside the core technology environment. The remaining 5 percent, however, can be decisive for competitive advantage. Another added, “Five years on, your competitors will be using agentic AI. If you are not, you will not be able to catch up.”
Standards, Governance and Interoperability. The absence of standards emerged as one of the most pressing concerns. Companies are building agents on different stacks, yet there is no shared rulebook to guide how those agents authenticate, exchange information, or take action across organizational boundaries. Leaders stressed that interoperability is not only a technical challenge, but also a strategic one. Without shared protocols, cross-enterprise interactions remain fragile.
Liability questions reinforced the urgency. If agents from two different companies take an action that leads to an error, who is responsible? As one participant put it, “We have no rulebook, but we are already playing the game.”
Balancing Autonomy and Human Oversight. Although participants welcomed the benefits of autonomous systems, they also emphasized the importance of oversight. All agreed that agents must remain subordinate to human authority, yet opinions varied on how much visibility is needed into an agent’s reasoning. Too much oversight reduces the value of autonomy, while too little increases risk. This balance between useful autonomy and trustworthy transparency will play a central role in shaping governance models.
An OpenText executive highlighted the growing evolution from human-agent collaboration to agent-agent collaboration. Agents are beginning to complete tasks with one another directly, which introduces new opportunities and new requirements for oversight.
Where Agentic AI Works Today, (and Where It Does Not). Despite ongoing challenges, the group pointed to clear use cases where agentic AI is already creating value. Software development is the most mature example. Agents can reduce backlog tasks, debug issues, and speed up development cycles.
Participants also described situations where current systems fall short. In e-commerce settings, for example, agents have provided incomplete or misleading inventory answers, leaving users feeling that no progress has been made. The consensus was clear. Agentic AI performs best when deployed to solve specific, well-defined problems.
In hospitality, one participant showcased a success story. Their agents gather customer information before a human employee steps in. This allows for more personal and efficient service. “We measure success in sentiment,” their CIO explained, noting noticeable improvement.
Cybersecurity leaders added that agentic AI is already essential in their domain. Autonomous agent swarms support proactive threat detection and rapid incident response. In high-velocity environments, autonomy is not optional.
Different Industries, Different Realities. While enterprise leaders described rapid experimentation, a municipal leader expressed a more constrained reality. Budget pressures, governance requirements, and citizen data sensitivity create significant hurdles. The contrast illustrated how uneven the adoption curve remains across sectors.
Architecture as the Next Competitive Advantage. As the conversation turned toward the future, architecture emerged as the defining theme. Participants cautioned against focusing too narrowly on individual use cases, which often leads to systems that cannot scale. Instead, they emphasized the need for reusable components, secure data pipelines, and governance structures that can support a broad portfolio of agents.
A Booz Allen Hamilton executive noted that building custom agentic architectures internally is increasingly impractical and distracts from core missions. Many around the table agreed that the industry is moving past the traditional build-versus-buy question. The competitive race now centers on identifying native, integrated platforms that provide autonomy within an enterprise-grade environment.
Others noted that their organizations are already seeing tangible economic benefits from agent deployments, further accelerating the push toward scalable, platform-based approaches.
This Bloomberg briefing was Proudly Sponsored By
——————————
Join the Conversation:
Instagram: @BloombergLive
LinkedIn: Bloomberg Live
Twitter: @BloombergLive
Interested in more Bloomberg Live virtual events? Sign up here to get alerts.
——————————