Why Most Companies Are Using AI Wrong (And How Winners Do

There's a growing chasm in the business [world, and it's not about who has access to AI](/article/solo-founders-million-dollar-ai-shops-reveal-what-big-companies-miss)—it's about who actually gets what to do with it.

A recent PwC study dropped a stat that should make every executive pause: three-quarters of AI's economic gains are being captured by just 20% of companies. And here's the kicker—it's not because those companies have better models or bigger budgets. They're playing a fundamentally different game.

The AI Daily Brief recently broke down what separates these winners from everyone else, pulling from PwC's research, McKinsey's AI Transformation Manifesto, and a fascinating case study from fintech company Ramp. The pattern that emerges isn't about tech specs or headcount. It's about something way more basic: most companies think AI is a tool for doing the same work faster, while the winners treat it as a platform for doing entirely different work.

The Productivity Trap

Here's where most organizations go wrong: they see AI as an efficiency play. Do more with less. Cut costs. Automate the boring stuff. And yeah, AI can do all that. But the companies actually winning? They're 2-3 times more likely to use AI for identifying new growth opportunities and 2.6 times more likely to say it helps them reinvent their business model.

McKinsey studied 20 AI leader companies and found they're seeing 20% EBITDA uplift on average, hitting breakeven in 1-2 years, and generating $3 for every dollar invested. That's not cost-cutting math. That's transformation math.

The difference shows up in where you aim the technology. McKinsey points to automotive supply chain integration as an example—companies like Toyota aren't using AI to make their existing supply chain 10% faster. They're fundamentally rethinking how supply chain integration works when you have these capabilities.

The Coordination Problem Nobody Talks About

George Sivulka from a16z nailed something crucial in a recent essay: while AI has made individuals 10x more productive, no company has become 10x more valuable as a result. Why? Because individual AI and institutional AI are completely different beasts.

Sivulka frames it with a thought experiment: "Imagine you doubled your organization's headcount tomorrow with clones of only your best employees. Each of these employees have minor differences, predilections, quirks, and perspectives. If they're not sufficiently managed, if they're not sufficiently communicating, if their swim lanes, OKRs, roles, and responsibilities are not well defined, you've created chaos."

That's what's happening right now in most organizations. Everyone has their own ChatGPT habits, their own prompting styles, their own workflows that don't connect to anyone else's. You've got thousands of people rowing in different directions, each individually productive but collectively generating noise.

What Winning Actually Looks Like

Ramp, the corporate card and spend management company, decided to stop waiting for vendors to solve this and built their own internal AI system called Glass. Their co-founder Eric Glyman tweeted that 90% of Ramp uses AI daily, but they noticed most people were stuck—not because the models weren't good enough, but because the setup was too painful.

What they built is basically organizational-level AI infrastructure, and the principles behind it challenge a lot of conventional wisdom about enterprise software. First principle: don't dumb it down. Seb Goden, who runs internal AI at Ramp, wrote that "the models are already exceptional, but most people use them like driving a Ferrari with the handbrake on."

Most enterprise software assumes non-technical users need simplified, dumbed-down interfaces. Ramp went the opposite direction: preserve full capability, just make the complexity invisible. Power users get multi-window workflows, deep integrations, scheduled automations, persistent memory, reusable skills. Everyone gets that on day one.

This breaks from how we usually think about software adoption, but it makes sense when you remember that AI itself is your coach. You're not figuring things out alone anymore—you've got the world's best tutor sitting there ready to help you work through problems. That changes what's possible for the average employee.

Second principle: one person's breakthrough becomes everyone's baseline. When a customer experience engineer at Ramp builds a Zendesk investigation workflow that pulls ticket history, checks account health, and suggests resolution paths, it goes into an internal marketplace they call "Dojo." Now the entire support team has that capability. They've shared 350+ skills this way, with an AI guide called "Sensei" that recommends the five most relevant skills based on your role and what you're working on.

Third: everything connects from day one. Glass comes pre-configured with all of Ramp's internal tools—one sign-in and you're integrated. When a sales rep asks Glass to pull context from a Gong call, enrich it with Salesforce data, and draft a follow-up, it just works. That's organization-level context engineering, and it's unsexy infrastructure that makes everything else possible.

The Build vs. Buy Question

Why would Ramp build this instead of buying something off the shelf? Goden's answer is straightforward: "Internal productivity is a moat. Using AI well is now a core business need." If AI competency is becoming a competitive advantage, outsourcing it means outsourcing your advantage.

McKinsey backs this up in their manifesto, arguing that 70%+ of AI talent should be in-house. This isn't about IT departments anymore. Senior business leaders who own different lines of business need to combine their domain expertise with AI know-how. It's a people transformation dressed up as a tech transformation.

The speed point is worth emphasizing too. McKinsey notes that the half-life of skills is shortening, and speed is becoming the defining organizational advantage. Companies that can absorb new capabilities faster and distribute them across their entire workforce are compounding advantages their competitors can't match.

What This Means for Everyone Else

Here's what makes this moment both exciting and terrifying: the gap between AI leaders and laggers is widening, not closing. The companies that figured out institutional AI early are building moats that don't depend on having access to better models—everyone has access to the same models. Their moat is organizational capability.

For companies still treating AI as a productivity tool you drop on people's desks, that's a problem. The coordination layer matters. The context engineering matters. The systems for distributing breakthroughs matter. You can give everyone Claude or ChatGPT and still be falling further behind if you're not thinking about how all those individual AIs coordinate.

The good news? The companies winning right now aren't doing black magic. They're building systems, investing in infrastructure, treating AI capabilities as strategic assets worth developing in-house, and fundamentally rethinking what's possible when everyone in your organization has access to powerful AI—properly harnessed.

The question isn't whether your company will adopt AI. It's whether you'll adopt it as a tool or as a transformation. Because increasingly, that difference is everything.

—Tyler Nakamura, Consumer Tech & Gadgets Correspondent