Accenture
Company Deep Dive
Accenture's AI bookings doubled. Its order book still shrank.
What a Company Deep Dive analysis finds in the filings, weeks before Accenture reports again
Accenture's generative AI bookings nearly doubled last year, to $5.9 billion. Its total new bookings still fell 1%. We ran a Company Deep Dive analysis to see which way the evidence points. Seven findings, summarised here from the full report, each with its source attached.
Picture: generated with AI for Dexil.Two confident answers circulate about what generative AI does to consulting, and they point in opposite directions. The first says AI eats the model from the bottom. Junior analysts do the research, the synthesis and the slides. That work automates first, and their billable hours pay for the partners above them. The second says AI is the largest transformation programme in corporate history, and somebody has to build it.
Accenture is the cleanest place to test this, for a reason that has nothing to do with its size. It publishes new bookings, the value of contracts signed before any of it becomes revenue. Most of its rivals are private partnerships and never disclose them. Bookings are what the argument is really about, because they show demand arriving rather than demand being worked off.
So we ran a Company Deep Dive on Accenture. That is one company researched end to end, and the output is a sourced report. The answer in the filings is narrower than either camp claims. Generative AI bookings nearly doubled in fiscal 2025, to $5.9 billion. Total new bookings fell 1%, to $80.62 billion, after rising 14% the year before. Both figures are Accenture's own. Together they say AI is not yet adding to the total. It is substituting for the work it displaces.
That reading has a date attached. Accenture reports its fourth quarter and full year in late September. If constant-currency bookings turn convincingly positive, the substitution reading weakens. If they do not, the softness starts to look structural rather than cyclical. Everything below was written before that print, and every finding carries the source it rests on.
What the analysis found
Business model
The AI line doubled and the total still went backwards
Two numbers carry the whole argument, and both come from Accenture's fourth-quarter release. Generative AI bookings reached $5.9 billion in fiscal 2025, up from $3 billion the year before. Total new bookings over the same year were $80.62 billion, down 1% in both US dollars and local currency. The year before they were $81.2 billion, a 14% rise. So the fastest-growing line in the business nearly doubled while the book containing it shrank. Whatever AI did to demand at Accenture in fiscal 2025, it did not add to the total.
Business model
In constant currency, the order book has been flat or falling all year
The fiscal 2026 quarters read better in headlines than in substance, because the dollar did some of the work. Second-quarter bookings were $22.11 billion, reported as a rise of 6% in US dollars. In local currency the rise was 1%. Third-quarter bookings were $19.32 billion, down 2% in US dollars and 3% in local currency. Strip out the currency and the order book rose one point, then fell three. Coverage tends to quote the dollar figure, so this is the finding a reader is most likely to miss.
Management quality
A one-point guidance change, a seventeen-point share price move
Accenture guided fiscal 2026 revenue growth of 2% to 5% in local currency when it reported fiscal 2025 in September. In March it raised the floor to 3% to 5%. On 18 June it cut the range to 3% to 4%. The top end came down by one percentage point, and the floor of the earnings guidance moved up. The shares closed at $156.01 on 17 June and $127.98 on 18 June, a fall of 17.9%. A move that size against a change that size is not the market repricing a quarter. It is the market repricing the model. One thing the fall on its own does not show: the stock closed the June gap on 28 July and reached $176.89 by mid-August, above where it started.
Technological
The exposure is the pyramid, not the demand
The analysis weights this judgement most heavily. The threat is structural and internal, not a question of whether clients still want the work. Consulting economics rest on leverage. A few senior people direct a large base of junior ones, and that base does the research, synthesis, first-pass modelling and slide production. Generative AI is good at exactly that work. If three associates and a model produce what ten associates used to produce, the engagement bills for three. Demand can hold up and revenue per engagement still falls. That same base is where a firm trains its future partners, so the effect reaches the leadership pipeline as well. This is reasoning rather than a reported fact, and the analysis labels it a forecast. It is here because it is the mechanism both camps are arguing about, and because the bookings numbers above fit it.
Porter's Five Forces
The model vendors have moved into implementation, as rivals and as partners
The firms that build the models now sell the deployment of them. OpenAI announced its Deployment Company in May 2026. It raised more than $4 billion from nineteen firms, led by TPG with Advent, Bain Capital and Brookfield as co-lead founding partners, and it puts OpenAI engineers inside enterprise customers. Microsoft announced its Frontier Company in early July 2026, with 6,000 engineers and salespeople and a $2.5 billion budget. Read one way, both cut the integrator out. The fuller picture is more mixed. OpenAI has signed multi-year deals with Accenture, Capgemini and McKinsey to deploy its platform, and Microsoft names Accenture among the systems integrators it runs these engineering partnerships with. So the same move creates a competitor and a channel at once. Which one dominates is still unsettled.
Legal
A federal investigation the company itself cannot size
Accenture Federal Services made a voluntary disclosure to the US government. The Department of Justice then opened a civil and criminal investigation. It concerns whether employees gave inaccurate submissions to an assessor checking federal security controls. Accenture is responding to an administrative subpoena and cooperating, and says it cannot determine when or how the matter will resolve, or estimate the likely cost. It can name the range of consequences, though, and does so in its own filing. Civil and criminal penalties under the False Claims Act. Termination of contracts, forfeiture of profits, suspension of payments and fines. Suspension or debarment from doing business with US government agencies. That is Accenture's own language, not a characterisation of it, and the disclosure is long-standing rather than new. It sits on top of a US federal business that Accenture estimates is taking about a percentage point off fiscal 2026 revenue growth. The June guidance cut pointed at the same exposure.
Valuation
The market is pricing the disruption, not the discount
The analysis carries a valuation lens from AFG, the Applied Finance Group, used under licence. AFG values a company on cash returns against its cost of capital rather than on accounting earnings. On its model year 2026 data, struck on 19 August, Accenture's market capitalisation of $105.8 billion sat 56.3% below an intrinsic value of $242.4 billion. It sat 53.6% below the consensus analyst target. Three-year total shareholder return of minus 18.9% a year puts Accenture in the bottom 12% of its peer set. Read on its own, that is a deep discount. Read against the findings above, it is the same story told in price. AFG puts 77.9% of the modelled value on future investment rather than on existing assets, and that is precisely the part the pyramid question puts in doubt. The gap does not show the market is wrong about Accenture. It measures how much of this disagreement is still open.
Run this on your own market
This analysis was produced on the same platform our customers use. You can open the full report in Dexil, and the free tier will get you through your own first one.
How this was produced
- Question submitted
- “The central question is whether generative AI structurally disrupts the consulting model or accelerates it. Analysts publicly disagree: one view is that AI automates the junior-consultant pyramid that underwrites billable leverage; the other is that AI expands transformation demand. Examine the FY2026 guidance cut, the sequential decline in new bookings, deals pushed to FY2027, the OpenAI and hyperscaler alliances, headcount and pyramid economics, managed services versus consulting mix, and DOGE / US federal exposure. Assess pricing pressure from outcome-based and AI-augmented delivery.”
- Analysis ID
- AccentureV4_CDDv3_2026-09-02_095853
- Executed
- 2 September 2026
Module overview
We asked the Company Deep Dive module the question above. It researches one company in depth: its filings and releases, what it says about itself, and what others say about it. It then works out which of those things bear on how the business performs from here.
The process is the same every time. It gathers many sources and extracts the individual claims each one supports. It groups those into distinct findings, then traces what follows from them through the business, first the direct effects and then what those effects cause in turn. Each step is scored for how strongly the evidence supports it, how much of the business it touches, and how soon it bites. The output is organised against established strategy frameworks, so one company can be compared against another rather than read as a one-off essay.
We commissioned this analysis ourselves, on a company we picked, using the same platform our customers use. No customer data went into it. Every finding was checked back against the primary document before it was published here, and each one links to what it rests on. One draws on a valuation dataset licensed from AFG and says so. It is still AI-generated analysis: a starting point for your own judgement, not a substitute for it.
See our AI disclaimer for the limits that apply to all Dexil output.