Accenture: Fratricide
For a hundred years, consulting firms have focused on one-off projects to implement new technologies and organisational structures. These projects have taken many forms, but the process of offshoring white collar labour stands apart because consulting firms have been able to capture ongoing managed services revenue after the implementation period finished.
That trade is now unwinding. The irony of AI implementation is that while it will require a lot of implementation, the opportunity for consulting firms to capture ongoing opex spend is limited; and indeed, many of these implementations are focused on eliminating offshore managed services. Half of Accenture’s revenue base will be eviscerated - and they will wield the knife themselves. Fratricide.
Frederick Taylor’s theory of ‘scientific management’ was, as Judith A. Merkle put it in her 1980 book Management and Ideology, "Taylor's system was an entrepreneurial scheme for selling organisational methods as science." To capitalise on his ideas, as he put it to Congress in 1912, companies would need to undergo a “total mental revolution” - and his men would help them do it. But while Taylor would visit a factory and make suggestions while charging for his time, he took on no ongoing responsibility for the runningof the business. Taylor’s mental revolution was profitable without the need for a long-term contract. Ironically, the most notable industrial reorganisation of the time was not led by a consultant. In 1920, Alfred P. Sloan began to reorganise General Motors into the ‘M-form’ multidivisional structure: separating strategic planning from tactical operations in divisions like Chevrolet, Buick, and Cadillac. While Sloan would of course lead the company for many years, the M-form reorganisation was a one-time change.
Once Sloan had set the precedent at GM, consultants could offer the same product to other companies: helping them navigate the period of change, and then leaving them to it. Books like Peter Drucker’s 1946 bestseller The Concept of the Corporation explained what Sloan had achieved, spreading the gospel of reorganisation across America.
By 1960, three firms specialised in implementing these reorganisations: Booz; Cresap, McCormick; and McKinsey & Company. Marvin Bower, who led the firm from 1950 and worked closely with Drucker, wrote in Perspectives on McKinsey that the firm “became known, exclusively to many, as reorganizers.” McKinsey would come to town - and then they would leave.
Once the American market became saturated, the consultants began to focus on selling American “know-how” and management techniques to Europe. By the 1970s, according to the historian Christopher D. McKenna, “McKinsey had, quite literally, decentralised most of the large companies in Europe.”
As the opportunities for decentralisation were exhausted, consultants engaged in other projects, with technology implementation developing as a product line as the century progressed. Companies lacking the know-how to take advantage of computer systems had consultants come in to install them. While firms of consultants might maintain “house accounts”, establishing long-term relationships with particular companies, the consulting work was a series of projects, each different from the last, rather than a load-bearing part of day-to-day operations.
The proliferation of IT and the internet enabled the white-collar offshoring wave that took root from the 1990s onwards. This wave of implementation, with consultants capturing processes and converting them into standard operating procedures that could be carried out by offshore staff, promised more revenue than just one-off project fees: consulting firms could sign long managed service contracts by spinning up offshore operations themselves. For the first time, consulting could be both capex and opex, implementing change and then owning the results indefinitely.
Capgemini’s headcount illustrates the importance of managed services to the major consulting firms: 180,000 of their 350,000 employees are in India alone. Yet the most useful case study is Accenture, since they are the only major firm that breaks their revenue down into consulting and managed services. It’s an even split: in 2025, Accenture generated $35.1 billion of revenue from their Consulting division, and $34.6 billion from their Managed Services division. The balance is exquisite.
And yet AI developments are destroying the managed services business model. The opportunity of implementing agentic systems has made the consulting and managed services divisions into mortal enemies. Accenture is starting to engage in fratricide.The basic thesis for AI in white collar work is that AI can do all the tasks you would have offshored - and more. To use the framing used by Decagon and Pace, the goal is to move from SOPs to AOPs - agent operating procedures. Tokens are cheaper and more scalable than employees; headcount reduction is rational, and AI implementations can realise it.
The problem of replacing SOPs with AOPs is firmly in the wheelhouse of the major consulting firms; once elite FDEs blaze a trail, the vast majority of AI implementation work will be done by these firms.
The claim I’m making here is that AOP implementation is hard enough to require consultants - it can’t just be done by the firms themselves - but easy enough for Accenture to scale the work horizontally across tens or hundreds of thousands of employees. With coding tools making software engineering easier than ever, the challenge is requirements and context gathering - human tasks which can be performed by a wide base of normal people.
That said, I would claim that implementation work is one of the most AI-resistant white-collar functions. The main challenge is context management - figuring out what information needs to go in the context window, and how to get it there; how to hoover up Chesterton’s Fences and know-how - the phronesis of organisational navigation. This means that you can’t easily hand the task off to AI, because the difficulty lies partly in politics, and partly in the things the AI model is structurally unaware of - indeed, in many cases it’s level 2 chaos. Many great startups are working on process mining, but my sense (as a professional!) is that the job requires a meta-skill that is fundamentally automation-resistant.
The managed services product line is in trouble. The point is not just that spend will fall as labour is replaced with tokens, but also that this spend will no longer be captured by the consulting firms. None of these firms are building world-class platforms of their own; nothing that can claim to rival Foundry (which allows Palantir to charge a significant markup on hyperscaler/foundation model costs). Their expertise is in people, orchestrating hundreds of thousands of employees and efficiently assigning them to projects; but in a world where fewer people are managing and evaluating agents, that’s the wrong specialisation. Even if the consultants are implementing AI, the AI opex will go to 1. hyperscalers 2. foundation models 3. Functionally-specialised application-level technology providers (like Decagon and Pace).
So Accenture Consulting is going to have to go out and kill - surgically excise! - its own twin. It’s fratricide to the tune of tens of billions of dollars.
If I were a consulting partner at Accenture, I’d be working through the CRM, telling Accenture clients exactly how much they’re spending on managed services, and exactly how brilliant it would be if they could replace that spend with AI. I’d be training as many people as I could, throwing bodies at the problem of passing the right information into context, while leveraging partner firms (OpenAI Frontier? Palantir Foundry? Decagon and Pace?) to figure out the right architecture in which to organise these efforts.
If I were a partner in the managed services division, I’d be desperately transitioning my business to fixed-fee work. I’d be looking for ways to capture licensed workflows - and I might also be selling training data (computer use?) to the AI labs. Revenues will decline - my job would be to manage that decline.
If I were on the board of Accenture, I’d look to divest the managed services division, selling it to someone less AI-pilled, with a higher estimate for future cash flows - maybe a rival BPO, maybe private equity. I’d recognise the division for what it is - a historical anomaly - and re-orient the business around one-off fee-for-service work, with a hard-nosed appreciation for the power of the AOP.