← Research

Who's getting f@?!'d by AI?

A read through the last few quarters of earnings calls for the companies admitting AI is cutting their revenue today — publishers gutted by Google's AI answers, IT-services firms baking "AI deflation" into guidance, call centers losing volume — and the far larger crowd denying all of it while switching to usage-based pricing anyway.

On earnings calls, AI is a tailwind. It is always a tailwind. Ask any management team what AI means for the business and you’ll get productivity, new products, a generational opportunity to serve customers better. Nobody’s revenue is going down because of AI. Everybody’s is going up.

Which is statistically impossible, so somebody is lying.

We read through the last few quarters of transcripts looking for the ones who aren’t — companies that got on a call and admitted, out loud, that AI is taking money out of their pockets right now. Not “we’re monitoring the disruption.” Actual, quantified, it’s-happening-to-us damage. They exist. There aren’t many of them. But they fall into a handful of very clear buckets, and the buckets are the story.

The publishers are getting it worst, and they’ve stopped pretending otherwise

If your business runs on free Google traffic, AI is currently eating you alive. Google answers the query with an AI summary, the user gets the answer, nobody clicks through. The clicks were the business. The clicks are gone.

Patient zero is Chegg, the homework-help company, which told investors its Google traffic “dropped by 50%,”1 watched revenue fall 42% in a year, started suing Google, and is now running its legacy academic unit purely for cash. It is the canary, and the rest of the mine is talking:

  • Future plc, the UK publisher, issued a profit warning and admitted it’s running an explicit “Google Zero” plan — organic sessions down 15%, programmatic down 17%, e-commerce down 24%.
  • IAC’s Dotdash Meredith has flagged the “softness… from traffic to our owned and operated sites” — its dot-coms — as a headwind running for a year and a half.2
  • NerdWallet described its own funeral in one sentence: “consumers increasingly turn to AI overviews and LLMs over traditional search, resulting in steep organic search declines.”3
  • Reach plc, the Mirror and Express publisher, lost roughly half its Google referral traffic, most of it through Google Discover.
  • Townsquare Media watched AI search-traffic loss cut one programmatic line in half — $5 million to $2.5 million — in a single quarter.
  • VerticalScope, which runs hundreds of enthusiast forums, posted monthly users down 23% and revenue down 16%, “driven by a broader AI-driven shift in online search behavior.”
  • NAHL Group said its personal-injury lead traffic is “down between 25% and 60%” as AI Overviews answer legal questions before the user ever reaches its claims funnel — “a real problem,” in management’s words.
  • Teads became the first ad-tech firm to say it plainly: premium publisher views down 10–15%, and “the decline is accelerating because of AI summaries.”4

Watch the arc, because it repeats: Ziff Davis spent its November 2025 call blaming a traffic drop on Google “rank volatility.” By February it had given up the euphemism — Tech & Shopping revenue down 18% “largely due to a drop in web search traffic,”5 with affiliate commissions down about $25 million year-over-year. The denial has a half-life. It usually expires around two quarters.

And the genuinely counterintuitive part: the killer is not ChatGPT. It’s Google’s own AI Overviews, strip-mining the exact search traffic Google spent twenty years teaching these businesses to depend on.

In IT services, “AI deflation” graduated from a fear to a line in the guidance

A year ago, “AI will compress our pricing” was an analyst’s nightmare scenario. Now the offshore services giants put a number on it and stick it in the forecast.

HCL Technologies is blunt about renewal deflation: of its top-10 contract renewals last quarter, CEO C. Vijayakumar said only five came with additional business — down from eight of nine the quarter before — casting the AI push as deliberate, in what an analyst on the call called “proactively cannibalizing the existing business.”6 Birlasoft explained the mechanism with almost suicidal honesty: “we have to commit to the savings and underwrite some of the savings right upfront,” which means “you see revenue deflation in the immediate and the medium term.”7 Hexaware sighed that “we called out the AI deflation last quarter”8 — it’s now baked into 2026 guidance. Endava conceded AI productivity “sort of erodes revenues” in its traditional work, and Expleo admitted it’s handing 20–60% productivity gains straight to clients up front.

A tier up, the big names are restructuring around the same force while choosing softer words. Capgemini guides its non-AI book to stay flat through 2028 because “a lot of it is going to be absorbed by the AI compression.”9 Cognizant is rebuilding its pricing into “AI-infused rate cards” — from all-human labor up to fully autonomous delivery — “because clients no longer accept all-human labor economics.” Globant was bleakest of all: “I’m not in a position to prevent cannibalization,”10 as it shifts from selling seats to selling token-based “AI Pods.” Amdocs says GenAI savings now pressure pricing “at essentially every renewal.”

Here’s the tell inside the tell: the mid-caps name the number, and the mega-caps still mumble “offset by volume.” The split runs right down the Indian IT sector — Infosys, HCL, and Hexaware admit the deflation while Wipro, Mphasis, and Persistent field the identical question and wave it off. Same industry, same AI, opposite candor. The honest ones are telling you today what the rest will confess next year.

Anything that sells human hours is losing volume from the bottom up

Two faster buckets, because they rhyme.

In call centers, the volume is already walking out the door. Teleperformance blamed an outright revenue decline in the Americas on language models eating its content-moderation and trust-and-safety work first. TaskUs said its largest client’s automation will keep pulling volume “per plan” through 2026. IBEX said one big client’s AI agents cut its call volumes 20% in six months. The contact-center software stack saw it coming: NICE confirms its customers are cutting agent seats and is pivoting to per-session pricing, and Five9 and PagerDuty are scrambling off seat-based models before the seats vanish.

In freelance and staffing, the cheap work goes first. Upwork was surgically precise about where the floor gave way: substitution is “really jobs that are $300 and less on the platform… the smallest, most transactional types of work.”11 Fiverr says the same — simple gigs down, complex ones fine. ManpowerGroup sees AI already denting the white-collar coder and call-center hiring it staffs for. And in media localization, ZOO Digital is eating realized price cuts of 10–20% wherever a client lets it use AI. The pattern across all of them: AI takes the easy, low-value work first and leaves the hard stuff alone. For now. Sleep tight.

The deniers, and why their pricing says more than their scripts

Here’s the part worth slowing down for. Around 130 companies were asked some version of “is AI eating your business?” and said no. Almost all of enterprise software sits here — Salesforce, ServiceNow, Workday, Adobe, SAP, Atlassian — swatting the seat-compression question away like it’s beneath them.

Now, none of this is secret. The denials are in the prepared remarks for anyone to read. And the pricing changes are in the press releases for anyone to read. The interesting thing is that they contradict each other. Almost every one of these companies is, at the same time, tearing out per-seat pricing and replacing it with usage-based pricing.

Sit with the logic. If your software is sold per employee, and AI agents are about to do those employees’ jobs, then per-seat revenue is the exact thing AI threatens. A company that genuinely believed AI was harmless would have no reason to rebuild its entire pricing model on a deadline. When a company’s words and its pricing point in opposite directions, the pricing is the more honest signal — because it costs something.

The most thoughtful voice in this camp isn’t a denial at all. Constellation Software’s Mark Leonard — the famously silent serial acquirer — thought the threat serious enough to convene a rare special call in September 2025 purely to talk about AI.12 Asked whether customers might use generative AI to build their own software and walk away from his, he called it “a great question, but… one that we have been confronting forever,” and answered with a defense rather than a denial: vertical software, he argued, is “the distillation of a conversation between the vendor and the customer that has gone on frequently for a couple of decades” — not something a language model rewrites overnight. His one concrete concession was structural, not lost customers: AI saddles a business that used to have almost no marginal cost with a new one, the compute and tokens it now has to pay for.

Others deny the impact but can’t quite hold the line:

  • ZoomInfo admits to real, multi-year seat compression — then pins it on the 2022 software hangover rather than AI.
  • Okta says it isn’t feeling a material headwind, and attributes its softer renewals to a lapped Covid cohort instead.
  • YouGov concedes that research consultancy will be “directly replaced for the most part by AI”13 and that its survey panels will commoditize — while denying any net impact.
  • HubSpot denies damage to itself, then notes that inbound and SEO leads are down 27% across its customer base from the shift to AI search. It sees the fire clearly. It just reports from the one building that isn’t on fire.

Japan got honest first

The most striking candor showed up in Japanese small-caps, which started addressing the threat before anyone made them. Global Information, a market-research report reseller, took a double hit — its search leads collapsed and its customers started doing their own research with generative AI — and reported segment revenue down 12.3% and operating profit down 27%. One software firm flatly announced that “SaaS is dead” and pivoted to becoming an AI vendor. JBCC, a systems integrator, is guiding its own integration revenue down on the explicit premise that AI moves software development in-house to its clients.

But the dominant Japanese pattern is pre-emptive denial: roughly a dozen small-caps now raise the “SaaS is dead” question on their own calls just to knock it down — a tell that the fear has arrived even where the damage hasn’t. The best of them are almost too honest about it. One legal-services portal conceded that light legal consultations will simply leak to ChatGPT. One cloud-services firm admitted it had cancelled some of its own software and replaced it with an AI coding tool — and then, without apparent irony, assured investors its customers would never do the same thing to it.

The one line to take with you

The consensus on earnings calls is that AI lifts all boats. The transcripts say something narrower and more useful. AI is currently eating three kinds of business: the ones built on free search traffic, the ones that sold hours of human labor, and the cheap simple work at the bottom of every market. A much larger crowd swears it’s fine — and reprices for the apocalypse anyway.

When everyone reprices for a problem they all deny, believe the repricing.

Notes

  1. “We built an incredible service, which we believe is the #1 service. The issue for us is that our Google traffic dropped by 50%. And so we weren’t seeing the necessary traffic to come in.” — Dan Rosensweig, CEO, Chegg, Q3 2025 earnings call, November 10, 2025. ↩
  2. “The softness we’re seeing from traffic to our owned and operated sites or traffic to, I’ll call it, dot-coms today… that has been a headwind for the last 1.5 years or so.” — Tim Quinn, IAC, at the TD Cowen Technology, Media & Telecom Conference, May 27, 2026. ↩
  3. “In 2025, we faced headwinds as consumers increasingly turn to AI overviews and LLMs over traditional search, resulting in steep organic search declines.” — Tim Chen, CEO, NerdWallet, Q4 2025 earnings call, February 25, 2026. ↩
  4. “It’s difficult to put a specific number on it… the decline is accelerating because of AI summaries and the changes in discovery. So I think it is impacting the traffic to those websites.” — David Kostman, CEO, Teads, Q3 2025 earnings call, November 6, 2025. ↩
  5. “Tech & Shopping’s revenues declined largely due to a drop in web search traffic, which had a meaningful impact on our affiliate commerce revenues… We believe we can contain the damage through alternative sources of engagement over time.” — Vivek Shah, CEO, Ziff Davis, Q4 2025 earnings call, February 24, 2026. One quarter earlier the same decline was attributed to “rank volatility.” ↩
  6. “Now the renewal deflation, I did call out last quarter — in 8 of the 9 deals we got additional business. This time… among the top 10, only 5 of them gave us additional business.” — C. Vijayakumar, CEO, HCL Technologies, Q2 FY2026 earnings call, October 13, 2025. An analyst on the call summarized the strategy as “proactively cannibalizing the existing business.” ↩
  7. “What happens in an AI deal is we have to commit to the savings and underwrite some of the savings right upfront, which is why you see revenue deflation in the immediate and the medium term.” — Angan Guha, CEO, Birlasoft, Q4 FY2026 earnings call, May 6, 2026. ↩
  8. “I was saying we called out the AI deflation last quarter. While there was mixed commentary from our peers, we did say there will be deflation and we budgeted for it in our numbers.” — R. Srikrishna, CEO, Hexaware, earnings call, May 7, 2026. ↩
  9. “We see a lot of it is going to be absorbed by the AI compression. So we say it’s flattish projected a few years out. And here, we’re talking to 2028.” — a Capgemini executive, Capgemini Analyst/Investor Day, May 27, 2026. ↩
  10. “I’m not in a position to prevent cannibalization. So I want that transformation to happen… as AI grows, we will keep on growing.” — Martín Migoya, CEO, Globant, Q4 2025 earnings call, February 26, 2026, answering how token-based “AI Pods” might cannibalize seat-based revenue. ↩
  11. “What we talk about when it comes to substitution… it’s really jobs that are $300 and less on the platform. So if you think about where AI can substitute human work, it’s really in the smallest, most transactional types of work.” — Erica Gessert, CFO, Upwork, at the UBS Global Technology and AI Conference, December 3, 2025. ↩
  12. “I think it’s a great question, but it’s one that we have been confronting forever… Yes, yes. My personal belief is that we all can’t see in the future.” Defending the moat, Leonard added that “vertical market software is the distillation of a conversation between the vendor and the customer that has gone on frequently for a couple of decades.” — Mark Leonard, Constellation Software special call on AI’s impact on software, September 22, 2025 (not a quarterly earnings call). ↩
  13. “That has always been unscalable. And like most forms of consultancy, it will be directly replaced for the most part by AI.” — Stephan Shakespeare, CEO, YouGov, H1 2026 results call, March 24, 2026. ↩