If AI Writes the Code, Are IT Services Firms Finished? — Pzena's View of the Other Face of 'Beaten-Down' Stocks

If AI Writes the Code, Are IT Services Firms Finished? — Pzena's View of the Other Face of 'Beaten-Down' Stocks

AI stocks are flying high these days, yet one industry is strangely going its own way: IT services companies — the ones that offer consulting, build systems, and even run operations on clients' behalf. The market's verdict seems to be that once AI can write code, these companies will no longer be needed. Value manager Pzena Investment Management, however, read things exactly the opposite way in its July 16 column. In naming Accenture, Cognizant, and Globant — companies the market has cast out as 'structural losers' — as second-quarter highlighted holdings, the firm wrote that these businesses are not victims of AI adoption but could instead become the ones helping AI spread. To borrow the original wording, the message is that these stocks are not trading at "prices premised on terminal decline."

Is that really true? I laid Pzena's original column alongside the three companies' most recent SEC filings and checked them one by one.

Start With the Price the Market Has Assigned to These Companies

The starting point of this post is the table Pzena presented as of June 30, 2026.

Source: Capital IQ, Pzena analysis (Exhibit 1 in the original, data as of 2026-06-30). According to the original footnote, the forward P/E is based on next-12-month consensus EPS, and ‘normalized EPS’ is Pzena's own estimate. In other words, it is worth keeping in mind that the last column is not a market consensus figure but a multiple Pzena built separately around “the level of earnings these companies are inherently able to generate.”

Still, the numbers themselves are stark. Globant trades at just 15% of its 5-year median valuation, and all three companies trade at 4.5x or below Pzena's estimated ‘normalized earnings.’ Pzena flagged this point, going out of its way to use the phrase “a price that assumes terminal decline.”

The Two Fears the Original Article Flags — and Pzena's Rebuttal

The first fear is a shrinking of the revenue pool (shrink the revenue pool). IT services has traditionally been an industry that bills for people's time, so if AI lets a project that used to take 10 people get done with 7, it seems only natural that customers would want to pay less. Pzena writes that this is already showing up in contract renewal pricing: annual price reductions that used to run around 3–4% are now approaching 6–7%. For reference, according to the original article, the IT services industry accounts for about 30% of global technology spending.

Pzena's rebuttal to this comes in two strands. One is a change in contracting models. Under time-and-materials billing, the more efficiency improves, the fewer billable hours there are, so all the gains flow to the customer; but under fixed-price and outcome-based contracts, the company that can deliver a defined result at a lower cost gets to keep part of the savings. The other is the historical experience that "as work gets cheaper, there is more of it." Offshoring and the cloud transition brought the same deflationary wave, and when unit prices fell, ROI rose and the backlog generated more work. The original article sums it up by noting that all three companies slowed in growth, but their revenue itself did not collapse.

The second fear is disintermediation. The scenarios: customers bring the work in-house, software vendors and AI labs move down into implementation, or AI-native startups displace the incumbents. Pzena concedes that "some of this will actually happen," starting with routine, well-defined work. But it points to limits: large enterprises have no desire to directly manage technology transitions amid their own messy realities; even the platform companies ultimately needed independent partners during the cloud transition; and startups lack the trust and business context needed to win large enterprises' complex work.

The most clear-eyed sentence in the original article is this one: "This does not mean every IT services company wins. The most exposed are the ones that merely rent out cheap engineering labor; the winners will be those that combine deep engineering expertise with client trust and a real understanding of the client's business." In the end, Pzena's case is not an optimistic "the entire industry survives," but a selection: that among them, these three companies will each survive in their own way.

Three Companies, Three Resolutions — The Original Article's Claims and the Latest Filings, Side by Side

Accenture: A Consulting-Led Seat at the Table, and Demand Yet to Arrive

The original Pzena article introduces Accenture as "the world's largest IT services company," emphasizing that its heavy weighting toward consulting-led transformation work means it is always at the table when clients decide on direction. The fact that revenue per employee is about $90,000—nearly double that of the largest Indian outsourcing firms—is further evidence that this seat commands a premium. On the other hand, the article notes, the stock has been cut in half over the past year, with June's lowered guidance and an expanded acquisition fueling the fear. Pzena's interpretation is this: clients are simply deferring large programs as money pours into AI investments themselves; the work hasn't disappeared, it has been postponed, and Accenture is positioned to capture it when it begins. It also adds the calculation that dividends and share buybacks return roughly 12% of market capitalization to shareholders annually.

So what do the subsequent filings show? On October 1, Accenture reported its FY2026 (September 2025–August 2026) annual results in an 8-K. FY26 revenue was $74.2 billion, up 6% in U.S. dollars (5% in local currency), exceeding the company's guidance (3–4% in local currency), and adjusted EPS rose 8% to $13.97. Annual free cash flow was $11.6 billion, and shareholder returns—including $7.5 billion in share buybacks—reached a record $11.5 billion. Fourth-quarter new bookings were $22.2 billion (+4%, book-to-bill 1.2), and annual bookings were $84.5 billion (+5%), so the bookings flow itself has not stalled. FY27 guidance also called for revenue growth of +3–6% in local currency, GAAP EPS of $14.39–$14.81, and at least $9.5 billion in shareholder returns. That said, the year-end cash balance of $12.8 billion (as of 2026-08-31), up sharply from $11.5 billion a year earlier, also means—as confirmed by a separate 10-Q (filed June 18)—that the company remains in a net cash position even after accounting for current debt of $113 million and long-term debt of roughly $5.03 billion. "Slowing revenue + expanding shareholder returns" is the face of this company today.

Cognizant: The Defenses of the Company Most Exposed to the Bear Thesis

Pzena introduces Cognizant as a company “more exposed to the bear scenario than Accenture,” because a large share of its revenue is delivery-and-operations in nature, leaving more room for AI to cut human hands. As its line of defense, the original article offers two things. First, about 60% of revenue comes from healthcare and financial services, and claims, payment, and risk systems are areas where failure is not an option—so they can't simply be handed over to AI. Second, the company's TriZetto platform processes about two-thirds of U.S. medical claims and is buried deep in the system itself. Pzena also explains that fixed-price and outcome-based contracts will cross 50% of revenue for the first time this year—meaning the company is escaping its status as a ‘hostage’ of the hourly billing model and intends to capture its share of the productivity AI creates. As a demand indicator, the article cited 12-month cumulative bookings up 11% year over year (per the original article dated July 16).

But set against the latest filings, the nuance changes. On July 29, Cognizant released its second-quarter 10-Q and earnings announcement. Q2 revenue was $5.481 billion, up 4.1% year over year (in constant currency); adjusted operating margin expanded 40bp to 16.0%; and adjusted EPS rose 4.6% to $1.37. Full-year 2026 adjusted EPS guidance was raised to $5.70–$5.82. However, the ‘12-month bookings +11%’ figure Pzena quoted had slowed to +5% as of this release (12-month bookings of $29.1 billion, book-to-bill of roughly 1.3x), and Q2 quarterly bookings declined 6% year over year. The quarter also included $84 million in restructuring costs under Project Leap. The balance sheet is more interesting. As of the end of June, cash stood at $1.038 billion with short-term investments of $13 million, whereas short-term debt was $33 million and long-term debt grew from $543 million to $1.527 billion—the footprint of drawing $1 billion on the revolver to fund $1.6 billion in share buybacks and $1.3 billion in acquisitions during the first half. Subtract total debt of $1.56 billion from cash plus short-term investments of $1.05 billion, and net debt comes to about $500 million. That isn't a heavy debt load, but it does mean the original article's wording shouldn't be carried over as today's facts.

Globant: 'Design and AI pods,' but stagnant revenue

The smallest and most digital-native of the three is Globant. Founded in Argentina and staffed with Latin American talent, it focuses on building consumer-facing services, streaming apps, games, and fintech experiences for clients like Disney and Google. Pzena argues that "design is still a hard area for AI," emphasizing the value of iterating with clients to decide what to build. The source also explains that the company is changing how it sells: "AI Pods," subscription-based teams that pair engineers with AI agents, are an early attempt to move away from hourly billing. It adds that revenue of roughly $85,000 per employee matches Accenture's and has grown every year since 2022.

Set against its filings, it becomes clear that this company is the test case for the story. According to its August 13 second-quarter results (6-K), revenue was $614.4 million, "roughly similar to the prior-year period," with an IFRS operating margin of 3.2% (1.0% a year earlier) and an adjusted operating margin of 13.2%, down from 15.0%. Third-quarter guidance calls for revenue of $607 million to $615 million, or -1.6% to -0.3% year over year. Full-year guidance of $2,428 million to $2,462 million—i.e., -1.1% to +0.3% year over year—makes this less a "growth slowdown" than a "growth standstill." Meanwhile, the pricing-model shift, the core experiment in Pzena's thesis, is most visibly underway here. CEO Migoya said annual recurring revenue (ARR) for the Glob.AI platform reached $52.8 million in the second quarter, up 61% from the prior quarter, and is expected to hit at least $110 million by the end of 2026, noting that "enterprises deploy AI Pods and pay for outcomes and usage rather than time." Anthropic, Vercel, and OpenAI were named as partners. Cash flow turned positive at $12.6 million for the quarter, versus -$2.9 million a year earlier; for the first half, the company described it as record cash generation. Cash plus short-term investments at the end of June stood at $168.8 million, while total debt as of year-end (20-F) was $367 million—so the company is in a net debt position, though the scale itself remains small.

Value Trap or Profitability Re-Rating — Testing the Case with the Latest Cash Flows

Pzena's conclusion runs as follows. It is true that AI reduces the labor these projects require and puts pressure on prices, and that is why these companies have gotten cheap. But all three are businesses that convert more than 90% of their earnings into cash and trade at double-digit FCF yields, and their conservative balance sheets — net cash at Accenture and Cognizant, minimal debt at Globant — buy them time to ride out a painful transition. Today's valuation pays you for the "risk of a rocky transition" while giving almost no credit to the possibility that these companies will remain at the center of enterprise AI adoption.

Running that claim against the latest filings, here is what it looks like. Cash generation itself is still alive. Accenture delivered $11.6 billion of free cash flow in FY26 ($9.3 billion over nine months even on an operating cash flow basis), and Cognizant produced $657 million of FCF in the first half — operating cash flow of $832 million minus capital expenditures of $175 million. Globant is smaller in scale, but its quarterly FCF swung from negative back to positive. That said, Pzena's line about "converting more than 90% of earnings into cash" mixes fiscal-year and half-year comparisons, so a precise match-up is difficult; in Cognizant's case, putting its first-half FCF of $657 million next to half-year net income of $1.3 billion, it is hard to assert "over 90%." The cause is overlapping restructuring costs and acquisition outlays, so stated precisely, Cognizant's current picture is "cash is still coming in, but capital allocation (buybacks and acquisitions) is going out the door faster."

The value-trap arguments deserve an equally cool tally. Cognizant's slowing bookings (+11% → +5%, and -6% in the quarter), Globant's zero-growth guidance, and the original piece's observation that even at Accenture AI budgets are cannibalizing other IT spending are all falsifiable claims hanging on the hypothesis that "demand is not disappearing, it is being deferred." To be fair, Pzena's side also has falsifiable evidence. The fact that the three companies' revenues have not collapsed even as contract price cuts accelerate to 6–7%, and the fact that Globant has begun to demonstrate its experiment of switching from hourly billing to output-based pricing in actual numbers, as ARR growth, both count as evidence that the simple equation "falling prices = revenue collapse" does not hold for now. That said, Globant's ARR of $52.8 million is barely 2% of its roughly $2.4 billion annual revenue, so it is too early to conclude that the transition is picking up speed.

Conclusion

That is about as much as can be confirmed from the original article and the filings. In his July 16 column, Pzena introduced Accenture, Cognizant, and Globant as second-quarter holdings, arguing that they could become beneficiaries of AI rather than its victims, and as of the end of June the three companies traded at less than half of their five-year median valuation and at no more than 4.5x his own estimate of normalized earnings. The results reported since then show Accenture's annual revenue and earnings growth along with record shareholder returns, Cognizant's margin expansion and raised EPS guidance, Globant's stagnant revenue and cash flow that turned positive, and even the early numbers on the shift 'from billing by the hour to billing by output.' Conversely, what cannot be verified is equally clear. Whether deferred demand will actually return, whether the profit share on fixed-price contracts will outrun the pace of price declines, whether Cognizant's bookings slowdown is temporary — these are questions time will answer, and Pzena's 'normalized earnings' estimates are themselves the judgment of an asset manager that no one can independently verify.

Personally, I think the appeal of this thread lies not in 'buy it because it's cheap' but in the question itself. If the number of people doing the work falls, will revenue fall too — or will the companies that get the same work done more cheaply capture the profits? IT services is the industry where that natural experiment opens first, and the three companies' quarterly reports will serve as the scorecard for this question over the next few quarters. Of course, the observation that valuations are cheap and the judgment that a rebound will come are two different things. Pzena himself wrote that "the task is identifying which companies will win," stopping short of a definitive answer, so it seems fitting that our place is to keep the original article and the filings side by side and simply watch.

References

• Pzena Investment Management, “Highlighted Holding: IT Services” (Second Quarter 2026, posted July 16, 2026) — https://www.pzena.com/americas/institutional-investors/insights/highlighted-holding-it-services-2q-2026/

• Accenture, FY2026 Q4 and full-year results 8-K (SEC filing date 2026-10-01) — https://www.sec.gov/Archives/edgar/data/1467373/000146737326000037/q4fy26earnings8-kexhibit.htm

• Accenture, Form 10-Q (quarter ended 2026-05-31, SEC filing date 2026-06-18) — https://www.sec.gov/Archives/edgar/data/1467373/000146737326000032/

• Cognizant, Q2 2026 earnings release 8-K Exhibit 99.1 (SEC filing date 2026-07-29) — https://www.sec.gov/Archives/edgar/data/1058290/000105829026000030/exhibit9916302026.htm

• Cognizant, Form 10-Q (quarter ended 2026-06-30, SEC filing date 2026-07-29) — https://www.sec.gov/Archives/edgar/data/1058290/000105829026000031/ctsh-20260630.htm

• Globant, Q2 2026 earnings press release 6-K Exhibit 99.1 (2026-08-13) — https://www.sec.gov/Archives/edgar/data/1557860/000110465926095916/tm2623064d1_ex99-1.htm

• Globant, Form 20-F FY2025 (SEC filing date 2026-02-27) — https://www.sec.gov/Archives/edgar/data/1557860/000162828026012910/glob-20251231.htm

Disclaimer: This post is informational content summarizing public filings, analysis, and the author's views; it is not a recommendation to buy or sell any particular stock.