The Separation: Most People Talk to AI. The 7.7% Who Put It to Work Are Building What Comes Next

In June 2025, Wix paid $80 million for Base44, a six-month-old software company with no outside funding and, for most of its life, exactly one employee. Founder Maor Shlomo reached $1 million in annual recurring revenue three weeks after launch—in large part by directing AI systems to do work that once required an engineering team. (TechCrunch)

Base44 is an outlier—talent, timing, and a hot market all played their part. But it points to a divide I have come to call the Separation.

The Separation is not the divide between people who use AI and people who don't. That gap is closing on its own—a majority of working-age Americans now use generative AI in some form. (St. Louis Fed) It is a difference in how work gets organized: between those who converse with AI—one prompt, one answer, a person carrying every step—and those who delegate to it, handing off multi-step work that keeps running while they attend to something else. Skilled conversational users get real value. The difference is that delegated work doesn't stop when you stop typing.

In practice, delegation looks ordinary. An agent screens the week's leads, drafts follow-ups on overdue invoices, and has the Monday report waiting before the coffee is poured—while a person still approves anything that leaves the building. Same tools, different posture.

How Narrow the Second Group Is

In a June 2026 AI Barometer survey of 7,675 U.S. adults, conducted with Prosper Insights & Analytics, 7.7 percent of Americans said they use agentic AI. Two-thirds had never heard the term, and executives reported agentic use at 2.7 times the rate of employees. (Channel V Media) Self-reported use is not the same as competence or results. But it shows how early this is—and that it is starting at the top.

The window is also short. Generative AI reached 39.4 percent of American adults within two years of ChatGPT's launch—roughly double the internet's pace at the same age. (NBER) Divides that once took a generation to form may now form in a handful of years.

What the Evidence Does—and Doesn't—Show

The research points in this direction. It does not yet prove it.

In a landmark field experiment, 758 Boston Consulting Group consultants using GPT-4 completed 12 percent more tasks, 25 percent faster, at 40 percent higher quality. But on a task just outside the AI's capabilities, consultants using it were 19 percentage points less likely to reach the correct answer than those working without it. (SSRN) For anyone delegating work, that is the whole lesson in one study: the gains are real, and so is the cost of unsupervised trust.

A 2025 paper by researchers at Oxford and Google DeepMind named the broader concern agentic inequality, noting that agents act as "autonomous delegates rather than tools"—so advantages scale through delegation itself. (arXiv) And the labor market is pricing capability: PwC's 2026 AI Jobs Barometer puts the average wage premium for AI skills at 62 percent, up from 56 percent a year earlier. (PwC) That figure measures AI skills broadly, not delegation specifically.

None of this proves that people who delegate outperform skilled conversational users. The Separation is my reading of where the evidence points, not a settled economic fact. But the mechanism is easy to see: each workflow you hand off frees time to build the next one. Advantages like that don't add. They compound.

The RiskForward View: What This Means for Business

The Separation is not only happening between companies. It is happening quietly inside them. Wharton's Ethan Mollick has documented "secret cyborgs"—employees multiplying their output with AI while keeping it from their managers. (One Useful Thing) In many organizations, the agentic minority already exists: unmanaged, unmeasured, and invisible to leadership. That is both an opportunity and an exposure.

The opportunity is that this no longer requires an engineering department. Mainstream AI platforms now ship with agents that take multi-step tasks, operate software the way an employee would, and run on their own schedule. What remains scarce is not access. It is leaders who treat these capabilities as an operating model rather than a curiosity.

• Bring the cyborgs into the open. Replace quiet, unsanctioned use with sanctioned tools, clear guardrails, and permission to experiment.
• Define delegation deliberately. Decide what agents may touch, where human sign-off is required, and how output gets verified—before habits form on their own. The BCG result is the reason.
• Measure output, not hours. When one person supervises the work of several systems, time at the desk misses what is actually being produced.
• Treat delegated authority as an accountability question. Handing a task to an AI system does not hand it the liability—responsibility generally stays with the business. What changes is where errors can originate, who is positioned to catch them, and whether existing coverage responds. Most insurance frameworks were not built for that question, which I examined in The Surging Blind Spot.

The same dynamic may play out between companies—agent-native firms compounding past incumbents whose legacy systems keep agents in a sandbox. That deserves its own analysis, and it will get one in a coming article.

What the Separation Is Not

It is not a claim that everyone else will be left behind, and it is not permanent: 40 percent of the same survey's respondents said they would delegate a daily task to an agent. Broadly available agents could even narrow gaps, as the BCG study hinted when its weakest performers gained the most. Which way it breaks depends less on the technology than on how quickly the majority chooses to cross.

Crossing Over

Crossing doesn't require rare genius. It requires a decision: to treat AI not as a better search box but as capacity to be directed—then supervised, checked, and reinvested. The tools are already on the device you are reading this on.

The Separation is less a wall than a door most people haven't noticed is open. Those walking through early aren't a different kind of person. They simply decided to delegate—and to stay accountable for what they delegated.