In fairness, we mark the other side of the trade. If these companies can show — with a reasonable degree of confidence — that they will hold their earnings and free cash flow, let alone grow them at a steady pace, then a genuine deep-value case opens up: a −2.3% terminal rate applied to a business that is in fact stable would be a clear mispricing worth owning. The survivors, in our view, will be those that truly pivot the model — embracing AI to enhance their offerings and then monetising them differently, most likely on an outcome basis rather than by the billable hour — while those that cannot will falter completely. At the current juncture we see little evidence of that pivot being made with conviction, which is why we remain firmly on the side of avoidance.
| Revenue (Q3) | Diluted EPS | New Bookings | Book-to-Bill | Stock Reaction | FY26 Rev Guide |
|---|---|---|---|---|---|
$18.7bn +3% LC · slight miss |
$3.80 +9% YoY · beat |
$19.3bn −3% LC · turning down |
1.0x at the waterline |
−18% on a beat · near 52-wk low |
3–4% trimmed from 3–5% |
The figure that mattered in Accenture’s print was neither the headline revenue nor the nine-percent EPS growth — it was the bookings line. New bookings of $19.3bn fell 3% in local currency and dragged book-to-bill to exactly 1.0, the waterline that separates a services business still growing from one that has begun to shrink. Bookings are the leading indicator and revenue the lagging one; for the first time this cycle the leading indicator has rolled over, and it has done so at the single best-run franchise in the industry.
Management did what capable management always does: it itemised the softness — a ~$100m Middle East hit, a ~1% federal drag, “a couple of” large managed-services deals slipping into FY27. Each line is individually defensible. But our job is to read the aggregate signal rather than recite the itemised excuses, and the aggregate signal is unambiguous: clients are hesitating precisely where engagements are largest, longest and most labour-intensive. That is not a collection of one-offs; it is the shape of the disruption beginning to arrive.
The tell within the tell is the price action. A compounder that grew EPS 9%, returned $8.2bn year-to-date and yields close to 5% still fell ~18% on the day — the market declining to pay for trailing fundamentals because it has lost faith in the forward model. A de-rating of that violence on a beat is not a valuation wobble to be arbitraged; it is the market re-underwriting the terminal value of the entire industry. On that, we are firmly on the market’s side — and we read Accenture, the highest-quality name, as the leading edge of a problem that worsens the further down the quality curve you look.
We frame this deliberately as a stance on the entire consulting and IT-services complex rather than a verdict on any one ticker. Accenture features heavily because it is the highest-quality, best-diversified operator in the group and therefore the cleanest read on underlying demand: when the strongest balance sheet and the broadest client base in the industry cannot escape a bookings decline and the first signs of deal slippage, the names with narrower offerings, weaker franchises and thinner balance sheets are by definition more exposed, not less.
The mechanism we describe intensifies as one moves down the quality curve — from the global integrators, to the Tier-1 Indian exporters whose economics are most levered to labour arbitrage, to the labour-heavy business-process outsourcers where AI agents can displace seats most directly. The specific companies named throughout this note are illustrations of an industry-wide dynamic and examples of where the pressure is already visible — not the limits of where we expect it to reach. Our caution extends across the complex.
- 1The consultant’s moat was complexity, and AI dissolves it. For four decades the industry monetised one asset: the gap between what enterprises wanted from technology and what they could build themselves. ERP, the web, offshore delivery, cloud migration, RPA — every wave was hard enough to deploy that it minted a multi-year integration annuity and an army of certified specialists to capture it. Complexity was the moat, and the consultant sat in it collecting the toll. Generative and agentic AI is the first wave engineered to collapse that gap: the interface is natural language, the documentation writes itself, and the implementer and the end-user become the same person. When a finance team can stand up an agentic workflow simply by describing it, the integrator is disintermediated from the build-and-run layers — exactly where the headcount and the revenue live. And unlike cloud or ERP there is no multi-year migration to staff: adoption is incremental, self-served and reversible, so even the transition revenue the industry has always harvested between platforms largely fails to materialise. This is a contraction of demand, not a deferral of it.
- 2The leading indicator has already turned. A services firm’s revenue is a mechanical product of bookings, conversion and contract duration. With Accenture’s book-to-bill at 1.0, bookings down 3% in local currency, and management itself flagging slower conversion, two of those three inputs are now working in reverse. We expect book-to-bill to print below 1.0 at the bellwether within one to two quarters as AI tooling cheapens and clients need materially less implementation labour per dollar of outcome. Crucially, a falling book-to-bill compounds: each quarter of sub-replacement bookings lowers the base off which the next quarter is measured, so the deceleration tends to accelerate rather than mean-revert — and the market is front-running precisely that sequence.
- 3Retiring the “advanced AI” disclosure tells you which way the number was about to break. Accenture introduced a standalone advanced-AI bookings-and-revenue metric in 2023 to prove it was an AI winner; it has now retired it, on the reasoning that AI is “embedded in everything.” Read that twice. “Embedded in everything” is their phrasing; “needs no specialist to implement” is ours — and the two describe the identical reality. Firms parade a metric while it flatters the story and quietly fold it away when the comparison turns against them; were advanced-AI bookings still visibly out-growing the erosion in the legacy base, that number would headline the deck, not vanish from it. Disclosure is a choice, and a choice made by a sophisticated, investor-savvy management team is itself data — the decision to stop counting was the loudest thing said all quarter.
- 4Deal push-outs are the mechanism, not the anecdote. This is the pillar we weight most heavily, because it is where the thesis turns into cash flow. In a deflationary technology regime the rational client shrinks commitment — shorter tenor, phased scopes, deferred mega-deals — because next year’s models will be cheaper and better than this year’s. That behaviour mechanically compresses backlog duration, degrades revenue visibility and strips the predictability premium out of the multiple. Accenture’s two large deals “slipping to FY27” are the first visible instance; we expect it to generalise. And duration, once compressed, is sticky in the wrong direction: clients that learn to buy in three-month increments rarely revert to three-year ones, so the visibility that underwrote the sector’s premium does not simply return when sentiment improves.
- 5AI deflation breaks labour arbitrage at both ends of the pyramid. The economic engine of the industry — and of Indian IT especially — is the people pyramid: a wide base of inexpensive juniors doing code, test, maintenance and process work, billed at a spread over cost and leveraged beneath a thin layer of costly seniors. AI attacks the base first and hardest, because the base is the standardised, automatable work; that removes the cheapest billable hours today and hollows out the apprenticeship that produces tomorrow’s seniors. The defensive pivot toward fixed-price and outcome-based contracts looks clever but, we think, accelerates the harm: once price is decoupled from headcount, the client’s first question is “where are my AI savings?” — and the honest answer hands the productivity dividend straight back to the buyer. The cruelest part is reflexive: the cheaper and more capable AI becomes, the stronger the client’s hand and the faster the spread compresses, so for the first time in the industry’s history time is the enemy of the model rather than its healer.
Read together, these are not five separate risks but one expressed five ways: the implementation premium the entire industry was built to harvest is being competed away by a technology the client can wield directly. Every prior platform shift widened that premium; this is the first that closes it. That is why we treat weakness in these shares as information to be respected rather than an opportunity to be bought.
None of this is theoretical. The repricing is already underway across the services complex, and the casualties cluster exactly where labour intensity is highest and genuine differentiation is lowest.
| Company | What Happened | Read-Through |
|---|---|---|
| Accenture ACN US | Q3 FY26 new bookings −3% LC, book-to-bill 1.0; two large managed-services deals slipped to FY27; FY26 revenue guide trimmed to 3–4% LC; stock −18% to near a 52-week low despite a 9% EPS beat. | The bellwether is cracking. Backlog momentum has reversed and the market is repricing the model, not the quarter. If the best operator is marked down on a beat, the read-across to weaker names is brutal. |
| Teleperformance TEP FP | Shares ~−75% from 2022 highs; FY25 revenue −0.7%; Specialized Services −9.3% LFL on a lost visa-processing (TLScontact) contract; founder ousted, a McKinsey operator installed as CEO; trades ~7× earnings. | The thesis run forward in time. The de-rate began when Klarna’s AI agent displaced ~700 seats; “cheap” at 7× has been a trap for three years because the multiple discounts a shrinking base, not a dip. |
| TCS TCS IN | Lost (parts of) its ~20-year Royal Bank of Canada technology mandate; ~150 staff to be rebadged to the winner — Accenture. ~48% of TCS revenue is North America. | Two-decade incumbency — long assumed the stickiest moat in the business — did not survive one procurement cycle. Anchor relationships are now contestable as banks rebuild for AI. |
Teleperformance is the lesson to internalise, because it is our thesis already played out. The market began pricing AI disruption into the BPO leader in early 2024, when Klarna revealed an OpenAI-powered assistant was doing the work of roughly 700 agents. Since then the company has insisted AI augments rather than replaces its half-million staff, reshuffled its portfolio, removed its founder and parachuted in a McKinsey CEO — and the stock has still surrendered roughly three-quarters of its value while screening “cheap” at every step down. A single-digit multiple on a structurally shrinking earnings base is not a floor; it is a slide with a handrail.
The TCS–RBC episode supplies the other half of the picture — and a useful complication. The mandate moved to Accenture, so on the surface a “winner” exists. But the point is not vendor A beating vendor B; it is that the contestants are now fighting harder over a pool of work whose value-per-seat is being compressed by automation. Intensifying churn is what a shrinking-value market looks like from the inside, and the winner of a deflating contract is still left holding a deflating contract. That a twenty-year relationship changed hands at all is the more important signal: nothing in this industry is sticky anymore.
The common thread is the one that matters. In every case the market has stopped paying for backlog, incumbency or dividend and started discounting a future in which the work itself is worth less — a thread that runs from the Tier-1 Indian exporters through the European BPOs to the global integrators. The names differ; the mechanism does not.
For the Indian majors the mechanics are starker than for Accenture. Their model converts a wage differential into operating profit by applying standardised human effort to maintenance, testing, integration and business-process work — the precise tasks large-language and agentic systems perform first, cheapest and around the clock. As AI absorbs the base of the pyramid, three things happen at once: utilisation-driven margin compresses; net headcount growth, historically the truest proxy for revenue growth, decouples from revenue and eventually turns negative; and the apprenticeship that produced senior delivery talent narrows. Across the sector, managements are already guiding to muted discretionary spend and defending margins through layoffs and pyramid “optimisation” rather than through growth — the unmistakable posture of an industry managing decline, not funding expansion.
None of this requires a dramatic collapse to matter for equity holders. A model that compounded high-single-digits for two decades slipping to low-single-digit or flat constant-currency growth, on a compressing margin and a de-rating multiple, is enough to produce years of dead money — which is exactly what a value trap delivers. Investors do not need to believe consulting disappears; they only need to believe it stops growing while the market keeps repricing its terminal value. Accenture’s push into platforms (~$9bn of acquisitions this year, anchored by the Dragos OT-security deal) and the mid-market (Accenture Edge) is the most intelligent attempt in the group to escape by buying recurring, less labour-linked revenue — but it is, by definition, an admission of the problem, and it does nothing for the lower-quality, more arbitrage-reliant names that lack the balance sheet to reinvent themselves.
The entire bull case now reduces to a single word: cheap. Accenture trades around 10× earnings on a ~5% yield, against a twenty-year history closer to 25×; the Indian majors are de-rating from the high-teens and low-twenties; Teleperformance sits near 7×. We are unmoved, because a multiple is only “low” relative to an earnings base you can trust. The defining feature of a value trap is that the denominator falls faster than the price, so the equity looks cheaper precisely as it becomes more expensive. The arithmetic is unforgiving: a stock at 10× that merely holds its price while earnings fall 25% has silently re-rated to roughly 13× — the buyer of “10×” now owns “13×” and a smaller company, having lost nothing on the screen and everything on the thesis.
This is why we refuse to anchor on yield or buybacks. Capital returns can defend a share price for a time, but they are a transfer of cash, not a cure for absent demand; a 5% dividend is cold comfort when the addressable hours behind the earnings are being automated away. Buying the de-rate here is, to us, catching a falling knife that happens to pay a coupon on the way down — and the coupon is funded by the very earnings the thesis says are at risk.
- 1 · Book-to-bill breaks below 1.0 at one or more bellwethers over the next one to two quarters — the first hard confirmation, not just commentary.
- 2 · Revenue growth fades to flat-or-negative in local currency through FY27 as weaker bookings convert and contract durations shorten.
- 3 · Margins are “defended” the wrong way — via layoffs, pyramid reshaping and AI-led delivery — buying time while confirming that the problem is demand, not cost.
- 4 · Multiples keep compressing as the debate migrates from growth rate to terminal value, and capital returns become the only visible support.
- 5 · The yield-and-buyback “floor” gives way, exactly as it has at Teleperformance — and the screen-cheap multiple is revealed as expensive.
- Avoid / underweight the pure-play consulting and IT-services complex. We are not initiating on valuation alone — a low multiple is not a catalyst.
- Use strength to reduce, not weakness to add. A constructive Investor Day or relief rally is an opportunity to lighten, not a reason to re-underwrite the thesis.
- Own the deflation, not the deflated. We prefer the model and infrastructure layer that captures the productivity transfer over the labour pools that AI displaces.
- What would change our mind: book-to-bill durably back above ~1.1, stabilising pricing, and AI-linked revenue re-accelerating visibly faster than the legacy base erodes.
.png)
.png)
.png)
.png)