In a world flattened by LLMs and AI agents, investing in people may be our greatest competitive advantage.

By 2025, 95% of SaaS companies had invested in AI-driven use cases. More broadly, McKinsey reported that 88% of companies (out of almost 2,000 surveyed) were using AI in at least one business function, up from 55% in 2023.

Already in 2024, Jared Spataro, Microsoft’s CMO of AI at Work observed that LLMs were becoming a commodity. “An LLM on its own, no matter how impressive, won’t deliver truly valuable results until it’s grounded in your company’s specific knowledge,” he said. “In other words, today’s breakthroughs will become tomorrow’s table stakes.”

Taken together, these trends should make you stop and think. The race to establish AI dominance may have already ended, not with a clear winner, but in a hundred-way tie between you and everyone else in your segment.

This isn't the first time a technology lauded as the ceiling of innovation ended up as the floor.

Commoditization follows a predictable arc. A new technology emerges and early adopters build genuine advantage around it. Competitors take notice, adopt the standard, and the differentiation collapses.

When Gillette introduced its multiblade Mach3 razor, it was a clear leap ahead of single-blade models. But competitors converged on the new standard, and Gillette's sales growth flattened, prices were squeezed, and the advantage evaporated.

The pattern repeats across every technology cycle. Early PCs were sold on raw processing power, a genuine differentiator when the machines themselves were novel. Within a decade, computing was a commodity, and the advantage had shifted to software, then to data, then to networks.

Having a website was a competitive edge in 1997. By 2003, it was a prerequisite The same arc played out with mobile apps, social media, and data analytics, each arriving as a strategic differentiator and eventually dissolving into table stakes.

As Columbia Business School professor Bruce Greenwald put it: "In the long run, everything is a toaster." All great innovations eventually become commodities, bought on the basis of price and nothing else.

As companies compete to replace ever-growing swaths of their workforces and workflows with AI, they are hitting a wall.

Less than six months after putting hiring on hold in favor of AI customer service agents, Klarna reversed its decision. Despite the reduced cost of AI chatbots, the quality was not high enough to actually replace a person.

According to a study by Harvard Business Review, using AI at work had unintended consequences for employees. Engineers, for example, spent more time reviewing and correcting AI-generated code, adding to their overall workload. Other workers noted that their habit of prompting multiple AIs at the same time while toggling between tasks broke their concentration and increased their sense of pressure at work.

AI and especially LLMs are trained on the accumulated record of how humans write and speak. It reflects our language back at us with impressive fluency. But fluency isn't understanding. What AI produces is a sophisticated average: plausible, pattern-matched, and largely interchangeable.

In other words, AI can sound authoritative, but it can't reliably produce anything that isn't a recombination of what already exists

Consumer perception of AI is mixed and very much in flux. But studies consistently show that people remain skeptical of AI-generated content, especially in advertising.

One recent study found that labeling an ad as AI-generated led consumers to rate it as less natural and less useful, even when the content was identical to ads labeled as human-made. Attitudes toward the ad became more negative, and willingness to research or purchase the product decreased.

As researcher Dr. Fabian Buder put it: "When people know that an advertisement was created by AI, they like it less."

To add a sense of scale, the study also reported that only 21% of respondents trust AI companies, and only 20% trust AI itself, revealing a significant gap between the people deploying AI and its intended audience.

The trust deficit runs deeper than advertising. A global study found that 73% of consumers actively avoid businesses that don't show empathy, and 71% believe AI cannot forge genuine human connections. More striking: three in five consumers say they only engage with companies that genuinely care about their needs, while nearly three quarters say they would take their business elsewhere if a company showed a lack of empathy toward their situation.

In 2025, researchers at MIT Sloan published a framework that provides a more optimistic look at AI in the workplace. Rather than asking which jobs AI will eliminate, they asked: what are people uniquely good at?

Their answer, the EPOCH framework, identifies five capabilities that AI cannot replicate: Empathy, Presence, Opinion/Judgment, Creativity, and Hope. Critically, the researchers push back on how we talk about these skills.

"We deliberately don't call these 'soft' skills," said Professor Roberto Rigobon. "A 'hard' skill, like solving a math problem, is comparatively easy to teach. It is much harder to teach a person these critical human skills."

Consumer research is where that gap becomes costly. AI can classify a customer comment as positive or negative, but most can’t reliably detect nuances. A statement like “that was brave of the brand” reads as praise in a sentiment model, whereas a human researcher would understand it to be a critique. This is what happens when AI is only reading the transcript.

GetWhy’s Research Agent now reads those nuances through video, analyzing the facial expressions, tone of voice, and body language of every interview in your library alongside the words. The pause before answering. The expression that tightened slightly on one word. The smile that didn’t quite match the words. The transcript captured what people said. The video captured what they meant. Now both are queryable.

This is the limitation that matters most for insight teams. AI solves the volume problem: how to talk to enough people, fast enough, across enough markets. The harder problem, the one that determines whether research actually changes decisions, is meaning. What did customers really feel about that product launch? What’s driving the behavior the data can’t explain?

Those questions have always been answerable. The signals were always captured in the footage. They were just waiting to be read.

Over 80% of companies say they feel competitive pressure from peers to speed up AI adoption — many of them investing before they have a clear picture of the value it will create. The irony is that this race to AI parity is recreating exactly the conditions it was supposed to escape. When everyone has the same tools, the tools stop being an advantage.

The numbers bear this out. According to a survey of 2,000 CEOs, only 25% of AI initiatives have delivered expected ROI, and just 16% have scaled enterprise-wide. At the same time, half of those CEOs acknowledge that the pace of investment has left their organizations with disconnected, piecemeal technology. They're spending more, integrating less, and breaking even on only one in four bets.

In a world where the tools have converged, humanity is the strategy.

"It has never been easier to sound like an expert,” said John Winsor in the Harvard Business Review. “When everyone can perform authority, authority itself loses meaning."

In a world where the tools have converged, the question is not whether to use AI. It is whether your AI is reading the full human signal, or just the easy part of it.