The Economics of AI
This post provides highlights from the AI Realized Executive Roundtable on June 25, 2025 in San Francisco. The focus was economics: how to measure value, justify investment, and scale AI initiatives without losing sight of impact.
This was the second roundtable in the series. The session was a closed-door small-group facilitated conversation among senior executives across pharma, insurance, cybersecurity, marketing, SaaS, and enterprise software. No slides. No spectators. Just peer-driven discussion, following Chatham House rules, about what’s working, what’s stalling, and what it’s really costing.
While every company is at a different stage of deployment, one theme rang clear: measuring its value remains a moving target. Here are a few signals from those conversations that stuck with us.
Cost Savings ≠ the End Goal
AI initiatives often start with headcount or cost reduction goals. But as one executive put it, “That’s just the wedge.” The real transformation happens when AI reshapes how work gets done. A pharma leader described how generative AI helped reduce clinical protocol design time from months to weeks and unlocked earlier market entry and revenue, not just cost savings.
“Efficiency? That’s just the wedge.”
Velocity Is a KPI
Across sectors, leaders are starting to see speed itself as a strategic advantage. From faster onboarding to accelerated content workflows, AI is shrinking time-to-impact. “When the rate of experimentation doubles, the business impact compounds,” said one participant. In this context, measuring velocity becomes just as critical as measuring cost.
“When the rate of experimentation doubles, the business impact compounds,” said one participant.
Different Metrics for Different Phases
Executives emphasized the need for phase-specific metrics. Foundational infrastructure investments often don’t show direct ROI, but they’re essential for future gains. Application-level metrics like customer satisfaction, usage per workflow, or content throughput can signal operational lift. As teams move into agent-based deployments, entirely new measurement models are needed.
Cross-Functional Benefits Are Hard to Capture
AI’s impact doesn’t fall neatly into departmental P&Ls. It spans functions, roles, and systems. A smarter onboarding experience, for example, benefits HR, sales, product, and customer success. But who gets the credit? Traditional ROI frameworks struggle to account for shared outcomes and emergent value.
No Standard Unit of Work
Unlike traditional automation, AI augments human work in complex, variable ways. This makes it hard to define a “unit of productivity.” Some leaders are experimenting with using AI to measure its own value by tracking tasks completed, time saved, or quality uplift but consensus remains elusive.
A Call for New Playbooks
The group reached strong consensus: organizations need new, adaptable frameworks for measuring AI value. One executive noted, “You’re measuring the past to change the future and yet the future is changing.” Without evolving metrics, businesses risk flying blind.
“You’re measuring the past to change the future and yet the future is changing.”
As AI adoption accelerates, these economic questions will only grow louder. That’s why AI Realized is doubling down on peer-led, real-world conversations like this one.