How the Best Companies Use AI

What It Covers

A PwC study, McKinsey's AI transformation manifesto, and Ramp's internal Glass platform reveal how top-performing companies use AI as a growth and business model reinvention tool rather than a productivity shortcut — and why building institutional AI infrastructure separates leaders from laggards.

Key Questions Answered

• •AI Economic Concentration: PwC research shows 75% of AI's economic gains flow to just 20% of companies. Those leaders are 2-3x more likely to use AI for identifying growth opportunities and 2.6x more likely to report AI enabling business model reinvention — not just efficiency gains.
• •Measurable ROI Benchmark: McKinsey's study of 20 AI-leading companies found AI-driven business transformations delivered a 20% EBITDA uplift on average, reaching breakeven within one to two years and generating three dollars of incremental EBITDA for every one dollar invested — a concrete benchmark to hold internal AI programs accountable to.
• •Institutional vs. Individual AI: Individual AI productivity gains do not automatically compound into organizational value. Without a coordination layer — shared outputs, defined roles, unified workflows — thousands of employees using AI independently create noise, not leverage. Building institutional systems to align individual AI use is a distinct and separate challenge.
• •Ramp's Glass Platform: Ramp built an internal AI workspace where every employee gets a fully configured environment on day one, with SSO-connected integrations, a marketplace of 350-plus reusable agent skills, persistent memory, and scheduled automations. When one employee discovers a better workflow, the entire team inherits it automatically through the shared skill system.
• •Don't Limit Employee AI Ceilings: Ramp's design principle rejects segmenting employees into basic versus power AI users. Because AI itself can tutor and coach anyone through complexity in real time, organizations should configure systems that give every employee access to full capability — not simplified, restricted interfaces based on assumed technical limitations.