AI Adoption Is Burning Cash While Incentives Ignore ROI

The AI spending spree is already hitting a wall, and the real story isn’t about technology--it’s about incentives, delayed payoffs, and who’s left holding the bag when the bubble bursts. The most dangerous moment in any boom isn’t the peak; it’s the first sign of doubt, when early adopters double down while everyone else starts asking for receipts. This conversation reveals that we’re already there: companies like Uber and Microsoft are burning through AI budgets with little to show for it, while IPOs for AI startups like OpenAI and SpaceX loom at a combined $4 trillion valuation. The hidden consequence? The same forces that drove the dot-com mania--executive incentives to sound innovative, VCs pushing hype over fundamentals, and CFOs lagging behind spending--are repeating themselves. For leaders in tech, finance, or any industry betting on transformation, this isn’t just a warning--it’s a playbook for spotting where systemic overreach meets real-world constraints. Those who see the pattern early won’t avoid disruption; they’ll position themselves ahead of the correction.

Why the Obvious Fix Makes Things Worse

You don’t need to be a technologist to see the pattern: when a new tool promises to do the work of ten people, companies rush to adopt it. The logic is airtight--cut costs, scale output, future-proof operations. But the reality, as Scott Galloway points out, is that AI isn’t saving money; it’s burning through budgets faster than the humans it was supposed to replace. Uber blew through its entire 2026 AI budget in four months. Microsoft is canceling Claude code licenses because they’re too expensive. At Nvidia, one executive admitted compute costs are now “far beyond the cost of employees.” This isn’t a minor miscalculation. It’s a systemic reversal of the core thesis behind AI adoption.

And yet, the incentives to keep spending remain powerful. Executives are rewarded for appearing innovative, not for delivering ROI. Mentioning AI on an earnings call today is like adding “.com” in 1999--pure signal, little substance. A recent MIT professor’s survey found that only 5% of AI projects could be tied to actual financial returns. The rest? Theater. The system rewards motion, not results.

"While I think that they're really intoxicated--it was like using AI as much as you can and talking about it in your earnings call--it's like adding dot com back in the '90s."

-- George Hahn, summarizing Scott Galloway

This is where consequence-mapping exposes the flaw. The immediate action--adopt AI aggressively--feels productive. It satisfies board expectations, pleases investors hungry for disruption, and positions the company as forward-thinking. But the downstream effect is a ballooning cost structure with no corresponding revenue lift. Over time, this creates a fiscal hangover. The companies that survive won’t be the ones that adopted AI fastest--they’ll be the ones that waited, learned from the early missteps, and moved when the tech actually delivered value.

The system responds. As CFOs start demanding justification, procurement gets involved. The narrative shifts from “we’re all in on AI” to “we need to scale back.” And when that shift happens, it won’t be led by a visionary--it’ll be a Fortune 500 company quietly cutting contracts, reining in experiments, and refocusing on profitability. The real kicker? The IPOs are still coming. SpaceX, OpenAI, and Anthropic are set to go public at a combined $4 trillion valuation. That’s not based on earnings. It’s based on belief.

Scott’s advice? “Sell everything.” Not as hyperbole. As a warning. When the gap between promise and performance becomes undeniable, the correction won’t be gentle.

The Hidden Cost of Fast Solutions

In Hollywood, the fear is simpler: AI will replace writers. But Netflix co-CEO Ted Sarandos offers a counterintuitive take--AI isn’t the threat; it’s the tool. He’s seen writers who once vowed to fight AI now use Claude as a “writing partner,” training it on their voice, style, and process. It bounces ideas, not scripts. It doesn’t replace the room--because the room is where originality happens. AI, by design, delivers the most predictable outcome. Storytelling thrives on the unpredictable.

"AI is not built to do that ever. I mean, the tool itself is built to give you the most predictable outcome possible--the antithesis of what we're trying to do when you make a big TV show or a film."

-- Ted Sarandos

This distinction matters. The immediate fear--job loss--is real. But the deeper consequence is misaligned expectations. Studios that try to replace writers with AI for low-budget, formulaic content (like Hallmark movies) won’t save much. Script costs are about 1% of the budget. The real efficiency gains are elsewhere: pre-visualization of complex stunts, which improves safety and reduces costly on-set errors. People die on film sets. AI helps prevent that.

But here’s where the system bends: the value isn’t in replacement. It’s in augmentation. The companies that win won’t be the ones that cut labor costs with AI. They’ll be the ones that use it to make their people better, safer, and more creative. The delayed payoff isn’t cost savings--it’s quality, speed, and reduced risk. And because that payoff takes time to materialize, most organizations won’t wait.

The pattern repeats: the fast solution (replace workers) feels decisive. The slow one (augment them) feels uncertain. But over 12--18 months, the augmented teams outperform. They adapt faster, innovate more, and retain institutional knowledge. The replaced ones? They leave gaps no algorithm can fill.

The 18-Month Payoff Nobody Wants to Wait For

Meanwhile, Scott Galloway has shifted his focus from AI to GLP-1 drugs--medications like Zepbound and Mounjaro that are changing how we treat obesity. Unlike most drugs, which work “on average,” these work for almost everyone. And patients like being on them. They feel better, not worse. They lose weight, reduce chronic disease risk, and in some cases, reverse conditions. Obesity is a node for over 200 chronic diseases. Fix that, and you change longevity, suffering, and healthcare costs at scale.

But here’s the catch: the drugs work. The system doesn’t. America spent $5.3 trillion on healthcare last year--$15,000 per person--yet lives shorter lives than peer nations. The problem isn’t innovation. It’s delivery. The healthcare system is structurally broken, optimized for billing, not outcomes. Even if these drugs are transformative, getting them to people requires navigating broken channels, insurer resistance, and hospital consolidation.

This is where the real test of systems thinking begins. AI’s ROI is being measured in quarters. GLP-1s could save trillions over decades. But no executive is rewarded for gains that far out. The incentives are misaligned. The system rewards short-term cost control, not long-term health. So while AI gets overhyped and overfunded, a potentially larger revolution in human well-being struggles to scale.

The irony? We’re pouring billions into AI to save labor costs while ignoring a healthcare transformation that could save lives and money--if we had the patience to build the infrastructure.

What Happens When Your Competitors Adapt

Scott’s reversal on Iran adds another layer: the danger of incomplete action. He initially supported military strikes to degrade Iran’s nuclear capabilities. The opening moves worked--navy damaged, missile systems depleted. But then the momentum stopped. No follow-through. No leverage. The IRGC, far from seeking a deal, now believes they can achieve “total and complete victory.” The signal of strength became a signal of weakness.

"Looking back on what has happened here, it is hard to make a rational case for how in any way, shape, or form this was a good idea."

-- Scott Galloway

This is systems thinking in geopolitics. Actions don’t exist in isolation. They provoke responses. The initial strike changed the incentives on both sides. But without a strategy for what comes next, it created a worse equilibrium. The same applies in business. Launching an AI initiative without a plan for ROI, cost control, or integration creates more problems than it solves. Competitors don’t stand still. They adapt. They exploit your overextension.

The lesson isn’t just about restraint. It’s about sequencing. Do the hard work upfront. Build the controls. Align incentives. Then move. Because half measures don’t buy time--they create vulnerabilities.

Key Action Items

• Audit your AI spending now. Over the next quarter, require every team using AI tools to tie usage to measurable outcomes. If they can’t, pause spending. This prevents budget bleed and forces accountability.

• Shift from replacement to augmentation. Invest in training teams to use AI as a partner, not a substitute. This pays off in 12--18 months with higher creativity, better decision-making, and retained expertise.

• Prepare for the AI correction. As CFOs demand ROI, early adopters will scale back. Position your organization to learn from their mistakes--don’t be the one selling the tools when the market turns.

• Rethink transformation timelines. Technologies like GLP-1s or enterprise AI won’t deliver in quarters. Build strategies that span years, not sprints. This creates separation from competitors focused on short-term wins.

• Build systems that survive incomplete wins. Whether in tech, healthcare, or geopolitics, half measures backfire. Ensure every initiative has a clear endpoint and exit strategy. Discomfort now--slower starts, more planning--creates advantage later.

• Act on anxiety. When health, career, or business issues weigh on you, move. Action absorbs anxiety. Make the call. Send the resume. Schedule the meeting. The moment you act, the fear begins to dissolve.

• Sell the hype, keep the tool. The AI bubble may burst, but the technology remains useful. When the correction comes, double down on use cases with real ROI--safety in film, diagnostics in medicine, efficiency in logistics. Where others retreat, you can advance.