CNBC: AI's Third Inflection — Agents Now Do Work

Why CNBC Is Calling This AI’s Third Inflection Point

CNBC’s Jar Dosa opens with a live demonstration: two prompts turn into a polished 10-slide executive deck — the AI takes over the computer, researches, writes, designs layouts, and delivers. No human touch. This isn’t a tech demo from a startup pitch; it’s mainstream business media declaring a new era.

The three inflections: “The first one was two years ago with ChatGPT, then about a year ago when the reasoning AI agents came out, and now with these agentic systems — agents able to reason, take task, and actually do work.” The shift from “AI that answers” to “AI that executes” is now visible enough for CNBC to build a 30-minute segment around it.

The SaaS bloodbath: The IGV software ETF dropped nearly 30% in the first two months of 2026. “It hit gaming, legal, insurance, trucking, cyber security. The carnage was indiscriminate. The same technology that was supposed to save software companies is now what is threatening to kill them.” When AI agents can do the work that software automates, the software itself becomes the middleman.

No-code builders are real now: People with zero coding skills are building software, websites, and apps — doing in an afternoon what used to take a team of engineers weeks. One person described building 15 projects in three months without writing a single line of code.

The safety paradox: Anthropic scrapped its core safety pledge, replacing hard commitments with “non-binding publicly declared targets.” The Pentagon is threatening to blacklist them for refusing to remove guardrails for autonomous weapons. OpenAI researchers are resigning. The companies that promised to govern themselves are abandoning those promises because their competitors did first.

A lawmaker who uses Claude Code: New York Assemblyman Alex Bores — author of the RAISE Act, the first major AI safety law — is himself a Claude Code user. He submitted a pull request to improve their iMessage connector. “I not only have used it, I’ve actually edited some of the code and submitted a potential change on GitHub.” A regulator who actually builds with the technology he’s regulating — rare and significant.

5 Signals from CNBC’s AI Disruption Report

• Agents are the third wave — ChatGPT (2024) → reasoning models (2025) → agentic systems doing real work (2026). Each wave accelerated faster than the last.
• SaaS is the first casualty — 30% drop in software ETF in two months. AI agents doing the work that SaaS automated means the software layer gets compressed.
• Safety commitments are collapsing — Every major lab pointed to competitors as the reason to lower standards. The voluntary pledge model has failed.
• Regulation is a $125M fight — A super PAC backed by Greg Brockman and Andreessen Horowitz is spending to defeat the first major AI safety law. The stakes: whether any guardrails exist at all.
• The capability gap is closing fast — Alex Bores warns: “Anything that you say it’s only so-so at right now, in six months, it’s going to be amazing.” The window for proactive governance is narrowing.

What This Means for AI-Powered Organizations

CNBC’s piece captures the exact moment AI agents crossed from “impressive demo” to “business threat.” For organizations deploying AI teams, two things matter: the capability is real and accelerating (agents doing multi-hour work reliably), and the regulatory vacuum means you’re on your own for safety standards. The companies that build trustworthy AI deployments now will have an advantage when regulation inevitably arrives — as Bores puts it, “putting a seat belt on the Lamborghini doesn’t really slow down the Lamborghini, but it saves lots of lives.”