From Proof of Concept to P&L: Operationalizing AI at Scale at the MIT Sloan CIO Symposium

By Barnard Crespi | May 25, 2026

At this year’s MIT Sloan CIO Symposium 2026, one of the most practical and candid conversations focused not on AI hype, but on what it actually takes to operationalize AI at enterprise scale and connect it to measurable business outcomes.

Moderated by Mike Grandinetti , the session brought together senior technology leaders from Waters Corporation , Celigo, and Travelers to discuss the realities of deploying AI inside complex organizations.

Mojgan Lefebvre, Executive Vice President and Chief Technology & Operations Officer, Travelers

Brook Colangelo, Chief Information Officer, Waters Corporation

Amy Farrow, Chief Information Officer, Celigo

The discussion moved beyond experimentation and into the operational, financial, governance, and organizational challenges that CIOs are now facing as AI adoption accelerates.

Key Themes from the Session

AI Must Be Tied to Business Outcomes

A recurring theme throughout the panel was that AI initiatives cannot exist as isolated technology experiments. Successful organizations are tying AI investments directly to measurable operational outcomes, cost reduction, productivity improvements, customer experience, and revenue impact.

Examples discussed included:

• Redesigning enterprise operating models instead of simply automating existing workflows
• Replacing manual service functions with AI-enabled agentic processes
• Using AI to reduce operational friction in customer service and claims management
• Reimagining how human expertise and machine intelligence work together

The panel emphasized that deployment is not the finish line. Value realization is.

AI Transformation Is Organizational, Not Just Technical

One of the strongest insights from the discussion was the recognition that AI transformation is fundamentally a company transformation initiative, not simply an IT project.

The panelists repeatedly highlighted:

• The importance of executive sponsorship
• The need for cross-functional product teams
• Cultural readiness and change management
• Training tailored to specific roles and workflows
• The requirement for a “growth mindset” across leadership

Several speakers stressed that organizations treating AI purely as a technology rollout risk missing the larger opportunity and may struggle to generate enterprise value.

Agentic AI Requires Discipline and Governance

The conversation also explored the growing interest around agentic AI and autonomous systems.

The panel cautioned against assuming that agentic AI is the answer to every problem. Instead, the speakers framed agentic AI as a design choice that should only be used where:

• Processes are highly complex
• Multi-step orchestration is required
• Human-driven execution introduces too much cost or delay
• Governance and escalation paths can be properly managed

The panelists also noted that poorly governed agents can create operational risk, unexpected AI consumption costs, and significant oversight challenges.

Data Architecture and Observability Are Critical

Another important theme was enterprise readiness.

Before organizations can operationalize AI at scale, several foundational elements must exist:

• Strong data governance
• Clear access controls
• Auditability
• Reversibility
• Identity and action traceability
• Real-time monitoring and observability

One particularly practical insight was the idea of “working backward from failure.” The question organizations should ask is:

“If something goes wrong, how would we determine why it happened?”

That framing becomes a powerful pressure test for enterprise AI architecture and governance design.

The Economics of AI Are Becoming a Major Executive Issue

The panel also addressed an increasingly important topic: AI economics.

As organizations move from experimentation into production-scale AI systems, token consumption, autonomous agent activity, and model usage costs are becoming real operational and financial concerns.

The speakers emphasized:

• Hard consumption controls
• Monitoring and observability
• Cost accountability
• Chargeback methodologies
• Governance frameworks tied to business value

The message was clear: operationalizing AI without understanding the economics behind it creates significant enterprise risk.

Takeaway

Perhaps the most important insight from the session was this:

AI operationalization is no longer about proving the technology works. It is now about whether organizations can redesign workflows, governance models, culture, economics, and decision-making processes fast enough to create sustainable enterprise value.

The organizations that succeed will likely be those that move beyond experimentation and begin treating AI as an integrated business transformation capability, not simply another technology initiative.

Panel Summary Based on Session Notes and Discussion Transcript.