AI Bringing Out
the Brooms

Adapt or Die. The Tape Has Already Decided.

Financial markets do not wait for anyone. They reward good ideas and put bad ones in their place — ruthlessly, and almost always ahead of the public conversation. It is all reflected in the price. If a company does not evolve, it gets left behind, and the tape will tell you so before the press release does. That is the only honest framing for what is happening to corporate America right now.

Agentic AI — software that does not just answer questions, but actually executes multi-step work end to end — is the most aggressive efficiency tool corporate America has had in a generation. By the close of Q1 2026, almost 80,000 tech workers had already been laid off, and roughly half of those cuts were attributed directly to AI and workflow automation.¹ Bloomberg models AI-related displacement reaching as high as 502,000 roles economy-wide this year. Goldman Sachs estimates faster-than-expected AI adoption could add as much as 0.3 percentage points to the headline unemployment rate before year-end.²

That is not a forecast. That is the broom already moving. The question for an investor, an operator, or a worker is not whether the sweep is happening — it is which side of the broom you are standing on.

Why Wall Street Is Cheering: A Plain-English Income Statement

To understand why share prices are responding the way they are, you only need to think about one piece of paper: the income statement. Strip away the jargon and a company's income statement is just three lines that matter:

Revenue is the money coming in the door. Expenses are what it costs to keep the lights on, and salaries are usually the biggest single bucket. Profit is what is left over for shareholders. The conceptual point is mechanical: profit equals revenue minus expenses. If revenue stays flat but expenses fall, profit rises. If a company can do the same volume of work with software instead of people, salaries shrink, profit expands, and the per-share earnings number Wall Street obsesses over goes up. That is the entire mechanism.

This is the inhumane part of corporate America that nobody enjoys saying out loud: a layoff announcement is, on the income statement, an immediate cost reduction. If management can credibly argue that the work still gets done — because AI agents are picking up the slack — then the market reads the announcement as a permanent margin upgrade rather than a temporary cost-cutting scramble. That is exactly why, when Block announced it was cutting roughly 4,000 roles and leaning into AI, the stock jumped the following day.³ Investors were not celebrating the lost jobs. They were repricing the company's future profit margin.

The income statement does not have a feelings column. It is a math identity. And math identities, deployed at the scale of the S&P 500's combined operating expense line, do not need permission from public sentiment to move share prices.

The Negatively Affected: When the Product Itself Was the Vulnerable Workflow

Not every company is being rewarded. The clearest losers are businesses whose entire product was a workflow that AI now does for free or for pennies. The market has been brutal in identifying them.

Chegg (CHGG) is the textbook case — the first major publicly traded company to be unambiguously hollowed out by a generative-AI competitor.⁴ The homework-help platform has lost roughly 99% of its market value in 39 months. A stock that traded above $113 in early 2021 was trading near $1 by April 2026, and roughly $14.5 billion of market cap was wiped out. Subscriber count fell 31% year-over-year by Q1 2025; revenue collapsed 49% year-over-year in Q4 2025. The company has cut staff in two waves — 248 roles in May 2025, then another 388 (about 45% of the remaining headcount) in October 2025. The strategic problem was simple and unsolvable: ChatGPT gives students the same answer for free, and Google's AI Overviews started serving those answers directly at the top of search results, choking off the discovery channel that brought Chegg new customers in the first place.

Teleperformance (TEP.PA) and Concentrix (CNXC) — the two largest call-center and business-process outsourcing players — have spent the better part of the past two years getting hammered every time a recognizable brand replaces human agents with a chatbot. Teleperformance shares fell roughly 29% in a single session in 2024 after Klarna announced a chatbot built on OpenAI was doing the work of 700 customer-service agents, and the stock has continued to grind lower — about 40% off its January 2024 peak by the time the dust settled.⁵ Concentrix dropped more than 12% on weak Q3 results and now expects only about 2% revenue growth in 2026. These are not bad businesses. They are good businesses whose single-largest product line is, by definition, the exact workflow agentic AI does best.

Salesforce (CRM) sits in the more nuanced category. The company has actually sold its own Agentforce product successfully — thousands of paying customers booked in Q4 alone — but the stock is still down roughly 27% year-to-date, because investors are asking the harder question: if Agentforce is good enough to replace human seats, what does that do to the per-seat licensing model that funded Salesforce's entire rise? When the product cannibalizes its own pricing structure, even a winning launch reads bearish. Press has labeled this stretch the "Great AI Scare of 2026," dragging down legacy software names from IBM to Salesforce.⁶

The Winners: Picks, Shovels, and the Operating System

On the other side of the trade, capital is flowing to companies that either build the infrastructure agentic AI runs on, or have figured out how to charge for AI work directly rather than for human seats. The pattern, once you see it, is unmistakable.

Nvidia (NVDA) remains the most obvious pick: up roughly 8% year-to-date in 2026, even after a brutal first quarter for the broader software complex. The reason is not hype — it is that hyperscaler capex on AI infrastructure keeps getting revised up, not down, because customers are now deploying agents in production rather than experimenting.⁷

The bigger surprise is the second-order beneficiaries. Micron (MU) is up about 59% year-to-date, riding an unprecedented memory-chip shortage created by AI data-center construction. Comfort Systems USA (FIX) — a name most generalists have never heard of — is up roughly 77%, because someone has to install the mechanical and electrical guts that keep all those data centers powered and cooled. When people say "sell the picks and shovels in a gold rush," these are the picks and shovels.

ServiceNow (NOW) may be the most instructive winner because of how it has restructured its pricing. Rather than charging per human user, ServiceNow has rolled out an "Agentic ACV" tier — customers pay for tasks completed by AI agents.⁷ That single move flipped the AI narrative from "our seats will shrink" to "our usage will grow as our customers automate more work," and the stock has clawed back nearly half of its Q1 losses on the strength of that repricing. Palantir (PLTR) rounds out the bench — not because it is the cheapest stock on the board (it is not), but because its Foundry / AIP platform has become the de facto operating system enterprises use to actually deploy AI agents against their real data. Selling the operating system in any technology shift has historically been one of the most lucrative positions on the board.

The pattern is the thing. Winners either build what AI runs on (Nvidia, Micron, Comfort Systems), or charge for the AI itself rather than for the humans it replaces (ServiceNow, Palantir). Losers either are the workflow being automated (Chegg, the BPO providers), or have a pricing model still tied to the human seats AI is removing (Salesforce, much of legacy SaaS). Once an investor learns to read the income-statement consequence of a company's pricing model in the AI era, the names sort themselves.

How Washington Is Adjusting (Slowly)

Government is moving, but at the speed government always moves — one cycle behind the technology. The Trump administration released "America's AI Action Plan" on July 23, 2025, paired with three executive orders.⁸ The Department of Labor has been told to stand up an "AI Workforce Research Hub," to redirect career-and-technical-education and apprenticeship dollars toward AI skills, and to fund "rapid retraining" for individuals displaced by AI.⁹

The honest assessment, as of this writing in late April 2026: most of the implementation deadlines have already slipped. Reporting from Axios in late April flagged that several agencies have failed to deliver on key steps that were due more than a month earlier.¹⁰ Brookings has separately argued that retraining programs — even when well-funded — have a poor historical track record of moving displaced workers into roles of comparable pay.¹¹ The policy intent is clear; the execution gap is wide; and the retraining-as-cure assumption may itself be the weakest link in the chain.

Policy historically arrives after the dislocation is already priced in. By the time a program meaningfully stands up, the workers it was designed to help have already routed themselves elsewhere — not because the program failed, but because human beings cannot wait two years for a federal apprenticeship pathway when the rent is due in thirty days.

Where the Displaced Workers Are Going

While the policy machine catches up, workers are doing what workers always do: stitching together income from whatever is in front of them. The data is starting to show it.

The Bureau of Labor Statistics' multiple-jobholder rate sat at 5.1% in March 2026.¹² That means roughly 5 in every 100 employed Americans are now holding more than one job at the same time. For historical context: that figure was 5.0% as recently as the end of 2013, and the long-run average across the past twenty-plus years has lived in roughly the 5.0–5.5% band, with a peak near 6.8% in mid-1995.¹³ The headline number alone does not scream crisis — but the composition underneath it is shifting. The growth is not in workers holding two W-2 positions. The growth is in workers holding one traditional job and a gig-platform side hustle.

The clearest evidence of that shift is on the gig platforms themselves. Goldman Sachs research shows gig hours rose the most in the cities where traditional payroll growth slowed — in other words, when the W-2 paycheck shrinks, the DoorDash app opens.¹⁴ About 20% of Americans who lost a job, lost pay, or had hours cut in the past year reported turning to a gig platform to make up the difference. DoorDash itself reported roughly 24% year-over-year revenue growth in 2024 with about a 67% share of the U.S. food delivery market — the platform is, in effect, absorbing the labor-market overflow.

That migration toward "irreplaceable" service work is rational. It is also a step down in earnings power for a large share of the people making the move. The gap — between the old salary and the new combined gig income — is the single most important number to watch in the back half of 2026, because it ultimately determines what consumer spending looks like and therefore what corporate revenue looks like.

The NAV View: Adapt or Die Is Not a Slogan

Markets are not moral. They are a pricing mechanism, and right now they are pricing a very specific belief: that agentic AI is a permanent, durable cut to the expense line of corporate America, and therefore a permanent, durable lift to corporate margins. That belief is not wrong. It is incomplete.

The same forces that lift the margin line are pressing down on the wage line of the American consumer who eventually buys those companies' products. The losers in this wave are obvious and being repriced in real time. The winners — the picks, the shovels, the operating systems, and the firms repricing AI work directly — are getting the capital. The displaced workers are routing themselves toward gig and service work at lower effective wages. And the policy response is, as usual, one beat behind.

The broom is moving. The dust is being repriced. The question is not whether to participate — participation is no longer optional once a generation-defining tool has reached the production phase of adoption. The question is whether you are positioned on the side of the income statement that the tape is currently rewarding, or on the side it is currently sweeping. The market has already decided which side is which. It is up to the rest of us to read what it is saying.