Even though the Wall Street Journal has “oh so reasonable” quotes from McKinsey spokescritters as to how “The Firm” is adapting to AI, the headline suggests otherwise: AI Is Coming for the Consultants. Inside McKinsey, ‘This Is Existential.’ The article has McKinsey officials make the key concession, that AI can replace a fair bit of what McKinsey has been doing, and the firm is working hard to find ways to incorporate AI.
However, the article does not hone in on the nature of the danger that AI poses to McKinsey, which is not to the need for expert assistance. Much of the work firms like McKinsey does involves non-public and often generated-during-the-study information and hence will be outside LLM text-hoovering. It is to the entire professional firm model. Its economics depend on substantial mark-ups of non-parter billed time over their cost.1 So the fact that clients will expect a smaller engagement team to answer a question, say, “How do we improve the profits of our European operations?” than before is very likely to dent profits. That’s even more important than you might think, since pay to partners is a big McKinsey recruiting point. Less lucre puts consultants like McKinsey at a further disadvantage relative private equity and Wall Street in recruiting and retaining staff. That in turn strikes at the very heart of McKinsey’s brand premise, its eliteness.
On top of that, unlike the dot-com era, where there was an explosion of corporate demand for professional advice to help cope with this Brave New World, McKinsey is not getting enough of a demand uplift to offset the impact of the AI expectation, that McKinsey deliver more with a lot fewer people, as in lower billings. The Journal tells us, “Advising on AI and related technology now makes up 40% of the firm’s revenue” which I suspect is as broadly defined as possible so as to make McKinsey sound as if it is at the bleeding edge. The piece even mentions that McKinsey has already shrunk its headcount from 45,000 in 2023 to 40,000 now.
Back to the impact of AI on the economics of consulting firms. Professional service firms typically bill out non-owner professionals at 3 times cash compensation or ~2.4 fully loaded costs (again this is dated, with fewer vacations and sick days but more costly health care, the typical “fully loaded” level may have changed). In my day through the 1990s, McKinsey seemed to hew to this norm.
Note that there are two ways to achieve “leverage”. One is the markup of non-profit-participant time. The other is the number of juniors compared to partners. Three juniors on average to each partner-type is a lot less leverage than a dozen juniors to each partner.
And McKinsey partners are acutely aware of the importance of leverage. I was retained by McKinsey after I had left the firm to work on a project to sell a money management firm. It was clear that either the senior or the second-tier partner would need to travel internationally and spend face time with prospective buyers. The senior partner grumbled in front of me to the junior partner: “This will be unlevered time. We need to bill accordingly.”
A story will show how these economics relate directly to McKinsey’s eliteness.
In the stone ages, MBAs were rare and only a few programs had much credibility in Corporate America (Harvard, Stanford, Chicago, and Wharton; the rest were as late as the early 1980s seen as a notch below, save perhaps Columbia and NYU for Wall Street jobs). So McKinsey merely by virtue of having top school MBA (who were also often the graduates of elite colleges) could field a team that had the psychological intimidation factor of meeting with client executives who (in nearly all cases) would never have been hired by McKinsey. McKinsey did also have a track record of being a propagator of new corporate practices from as early as the 1930s.
In the 1980s, when Wall Street profits and pay levels exploded, McKinsey found that it was losing out in B school and campus recruiting for top candidates far more often than it had in the past to the likes of Goldman and Salomon. Even worse, a decent number of mid-career McKinsey professionals were leaving the firm to go to finance firm for 50% to 100% increases in pay. This was particularly destructive since those 3-6 years employees were engagement or senior engagement managers and were the working oars of the firm.
In 1987, I was brought in as one of only two non-tenured partners to assist three very senior directors, Carter Bales, Fred Gluck, and Don Waite, all of whom were contenders in the upcoming election for firm Managing Director (Gluck later became Managing Director). The other participant was Purnendu Chatterjee, then a non-tenured partner (aka “principal”) who left McKinsey to run a fund for Soros and now has his own investment firm). They had teamed up to address the question of what to do about the losses in “talent” that McKinsey was suffering. They deemed this to be so important that they did not want to tout different approaches in their campaigns but to agree on what to do to help the firm cope.2
The result (as I was told later) was that the firm decided the way to compete was to raise partner pay, and in a marked way someone “made partner” as in was elected to principal, which then was at the five and a half to seven year mark.3 The way the firm made the money math work was to increase the size of the personnel pyramid under the directors and principals, by more leverage of each partner (more juniors per senior) via staffing on studies and by lengthening time to principal and director.
The article mentions the current leverage level but in such a way that it’s still a bit opaque. From the Journal:
Traditionally, a strategy project with a client might require an engagement manager—essentially, a project leader—plus 14 consultants. Today, it might need an engagement manager plus two or three consultants, alongside a few AI agents and access to “deep research” capabilities, Smaje said. Partners with decades of experience might prove more indispensable to projects, in part, because they have seen problems before.
In my day, there were two levels of “manager”: an engagement manager (someone with at least 2-3 years of experience) and a “senior engagement manager” which would typically come at more like the 4+ year mark.
An engagement manager would lead 2-4 people, which would be a combination of more junior consultants and analysts (college grads).
It would only be a senior engagement manager that would have been in charge of a team as big as 14, and then with some engagement managers supervising the typical 2-4 professional on discrete pieces of work.
One person supervising 14 people with no intermediate minders is a train wreck, quality-wise.
In keeping, by the early 2000s, I was regularly hearing complaints from top search firms (the sort that does CEO/C-level searches) that the quality of McKinsey work had declined and was regularly out of line with the price. The same search firms were also often describing McKinsey as arrogant and aggressive, something that would have been unheard of with clients in my era, even by the most full-of-themselves directors. McKinsey was often selling (in my day, they instead marketed a ton and relied on referrals and follow-on work to generate their business) and seemed to be inept about it. I heard this sort of account more than once:
McKinsey professionals call on possible or even current client executives. They pitch a particularly engagement: “We think you should hire us to do X and here’s why.”
Client politely says they don’t think they need X done, or alternatively, they can handle X on their own.
McKinsey team gets aggressive: “No, you really need us to do X.”
This combination of factors no doubt contributes to the overwhelmingly negative comments about McKinsey in the Journal’s comment section, even before factoring their rap sheet of ethical lapses, including Enron, Purdue Pharma, UnitedHealth, cuts in patient service at NHS, bribery in South Africa, a Twitter crackdown in Saudi Arabia that may have played a role in the bone-sawing of Jamal Khashoggi.
Some excerpts from the Journal’s article:
Companies pay dearly for McKinsey’s human expertise, and for nearly a century they have had good reason: The elite firm’s armies of consultants have helped generations of CEOs navigate the thorniest of challenges, synthesizing complex information and mapping out what to do next.
Now McKinsey is trying to steer through its own existential transformation. Artificial intelligence can increasingly do the work done by the firm’s highly paid consultants, often within minutes.
That reality is pushing the firm to rewire its business. AI is now a topic of conversation at every meeting of McKinsey’s board, said Bob Sternfels, the firm’s global managing partner. The technology is changing the ways McKinsey works with clients, how it hires and even what projects it takes on.
And McKinsey is rapidly deploying thousands of AI agents. Those bots now assist consultants in building PowerPoint decks, taking notes and summing up interviews and research documents for clients. The most-used bot is one that helps employees write in a classic “McKinsey tone of voice”—language the firm describes as sharp, concise and clear. Another popular agent checks the logic of a consultant’s arguments, verifying the flow of reasoning makes sense.
Sternfels said he sees a day in the not-too-distant future when McKinsey has one AI agent for every human it employs.
An aside: I am a little surprised about the “McKinsey voice” part. How much “voice” is there in slide presentations? Again, in my day, where that showed up was proposals (written by partners) and articles in McKinsey publications, where McKinsey’s very skilled editors would whip a draft into shape. I was just about the only person evah to write a “progress review” as a text document (with charts and images in the text as well as in appendixes) in what was called “final report” format.4 Or are young consultants these days so poor at writing generally that they need AI to produce professional-sounding e-mails?
Mind you, AI will not kill consulting but it seems clear a lot of adaptation is in store for McKinsey and its big brand-name competitors, Bain and BCG5:
One immediate change is that fewer clients want to hire consulting firms for strategy advice alone…
Companies, [Nick] Studer, [CEO of Oliver Wyman] added, “don’t want a suit with PowerPoint. They want someone who is willing to get in the trenches and help them align their team and cocreate with their team.”
At McKinsey, Sternfels is trying to cement the notion that the firm is a partner, not adviser, to clients. About a quarter of the company’s work today is in outcomes-based arrangements: McKinsey is paid partly on whether a project achieves certain results.
Perhaps I am being unduly negative, but I see a lot of whistling past the grave here. Clients will expect much smaller and more senior-heavy teams, which throws a wrecking ball into how McKinsey generates enough revenue to produce the lofty pay to which its top brass has become accustomed.
Moreover, “smaller and more senior” increases the viability of boutique firms as competitors.
It also means that the ability to develop or access non-public information will become more valuable. A colleague who was an executive at an expert network firm that had two of the top three consulting firms as clients noted:
Expert networks will continue to be relevant to these firms. A useful function of expert networks that will not be replaced by AI is finding people who can offer the kind of on-the-ground perspective that doesn’t always make it into formal research or the business literature. These people provide information that is not secret, but not public. In that consulting firms are presumably paid to find and fix problems that have been overlooked, it is a relevant function in that AI simply scans information that is already public.
This “not secret but not public” information is far more widespread that one might think, and is particularly valuable to companies trying to enter a new market or second-tier players. In my day at McKinsey and after I started my own practice, I spent a substantial amount of my time generating that sort of information.
The expert network exec felt compelled to add:
I have noticed that a lot of people are treating AI-scanned documents as the letter of the law. I recently observed that the version of a law offered by ChatGPT was not only not up-to-date, but that the revised version of that law was already more than 15 years old and easily found on the Internet. I happened to be familiar with that law and its revision, but the person who sent me the GPT-regurgitated version was not. It was slightly disturbing that initially, I got a reflexive argument about it from a well-educated professional.
AI will absolutely not obviate the need for fact-checkers versed in solid research practices, if only to combat the widespread and questionable tendency, which I first observed when Wikipedia was newer, to treat Internet resources in the same manner as traditional sources.
Pundits, journalists, and our readers (particularly IM Doc!) have offered similar warnings about glaring errors in AI results, which nevertheless are too often fed into decision-processed despite that. How bad the information pollution becomes before clean-up or prophylactic efforts are made is anyone’s guess.
By contrast, it seems a safer wager that McKinsey will find it hard to preserve its current stature and profit level.
____
1 Keep in mind that even though the man who built McKinsey, Marvin Bower, modeled it on law firms, McKinsey does not provide periodic invoices of actual time expenditures, billing rates, and expenses. The practice in my day, and I doubt it has changed much, is that McKinsey engagement letters would identify the top members of the engagement team (as in the so-called Director of Consulting Services on that project, plus the non-tenured “partners” on it (recall that McKinsey is not a partnership) which then were called Engagement Directors. It would then give the size of the team that would work under them and the billing rate per month, plus the expense percentage assigned (again in my day, 15% for domestic engagements and 25% for international).
The result too often was that studies were “undernegotiated”, that the partners had promised the client too much in relation to the size of the team, which resulted in the worker-bees putting in backbreaking hours. The result too often was that promising consultants on those assignments resigning after they had landed a new job.
Mind you, the dot-bomb era led to a collapses in demand for consultants. McKinsey’s headcount in its North American offices was reduced by 50% over two years. So McKinsey may have also wound up shifting its engagement terms since my day, and particularly as a result of this huge demand downdraft.
2 It was immediately clear that these three had never been in a room together without much in the way of other adults present. The jousting was impressive to behold. I kept my head down and (without having been asked) gave them a primer on Wall Street economics, describing them by major line of business.
3 If a consultant did not make partner in that window, the firm had an “up or out” policy, much like law firms for associates who do not make partner.
4 I was shocked when I came to McKinsey from Goldman to see that clients accepted mere sideshows as the end product, since in six months, the client would have large gaps in the information conveyed. I insisted on putting a disclaimer on all of the front pages of my progress reviews something like “This is not a complete record of the presentation” (this was actually in the firm’s lexicon but bizarrely rarely employed). I was used to professional documents being in text, such as memoranda from our counsel, valuation letters, and analyst reports, for the avoidance of doubt, as the lawyers would say.
5 Bain will probably fare the best. First, it has a ton of captive business from its private equity fellow traveler Bain Capital for due diligence on possible investments. Second, Bain has always worked in a very hands-on way, sending in large teams to act in a managerial capacity. How that works from a liability standpoint of the company subject to a quasi-Bain takeover is beyond me
“Or are young consultants these days so poor at writing generally that they need AI to produce professional-sounding e-mails?” Yes, even the US equivalent of Oxbridge graduates.
It was “yes” for years even before xGPT, even at the Oxbridge-Ivy level.
AI has both made it easier to hide such deficiencies, and more obvious as it’s markedly noticeable when someone’s prose sounds like Stepford-GPT (to me at least, IMO)
What is it about writing from recent college grads that is so poor? Don’t get me wrong, I know their writing skills are poor. However, the quality of the writing I get from UG research assistants is not necessarily worse than what I remember reading from my classmates 20 years ago.
Are younger writers now much worse than young writers several decades ago? If so, is there something in particular that is off? I would like to help my undergrads become better writers but outside of “write a lot” I don’t really know what else to tell them. While “write a lot” is part of the equation, so is getting feedback, especially helpful feedback. A lack of helpful feedback has always been one of the main frustrations I’ve had while trying to improve my own writing skills.
It’s the curriculum.
There are no more “paleo-English” teachers who use/inspired by a book of style. The public school curriculum has de-emphasized skills like rote spelling. (I get it….rote-ness doesn’t equate greatness, but learning to spell helps recognized the ties between words, IMO).
Incessant use of passive voice—somewhere that’s been taught as being ok.
I can moan on…but that is a starting point (assuming Common Core has some hand in this, but not total fault). This is happening irrespective of wealth, spending, resources of the public school district. And most parents don’t bother to raise a voice cuz they don’t know better or just trust the school to do the right thing. (or like me, this ain’t the hill for me to die on when it comes to parent-teacher relations, lol)
Yes. They don’t read much.
But I do agree about the feedback part. In college, the thing that helped me most was my sophomore tutorial, where we wrote papers most weeks. Each of my two tutors would often give more in comments than I had written. It was painful (even with my doing well in the course) but it taught me to “see” on a sentence level what my writing actually said, as opposed to what I thought it said.
Feedback. Yes! When I took English Composition as a freshman we wrote an in-class paper every Friday. Usually filled a Blue Book on the subject of the week. Same thing in required English Literature classes. It was possible to spend a week on The Waste Land or Wise Blood or Walden, and it was expected that the works had been read. More than once a student was “excused” from class. That would give the Office of Student Affairs heartburn these days.
We were on the quarter system, which I have always preferred to the semesters – anything can be endured for 10 weeks. Anyway 10 weeks, 50 minutes per class period, M-F. 10 in-class papers plus two larger assignments. One at the midpoint and one as a final exam. With lots of red ink, even when you were doing well. In my very first college class (European History after Napoleon we read Fathers and Sons, Father and Son (Gosse), The Reason Why (Crimean War), August 1914, and The Rise and Fall of the Third Reich. The survey textbook was a supplement. Exams were discussion questions to be answered in the Blue Book. It was a different time. And in the age of AI, I expect the Blue Book to make a comeback. Along with handwriting. Nah, just dreaming…
A friend who is currently a professor of economics told me all of her classes are going back to the blue books. No laptops in class, no recording of lectures. If found to be recording a lecture, you are removed from class. There is hope.
“They don’t read much.”
I believe that this is the root cause. How well would one speak without rarely ever having listened?
“Read a lot” is also helpful. Just reading really good writers (the ones with a sense of rhythm, to begin with, and of the total value and relative weight of words) sets off a process of unconsciously absorbing those qualities.
But I gather most of this discussion centers around utilitarian writing how to convey information clearly, concisely, accurately. That begins with thinking. I suspect, though cannot prove, that most who don’t write well don’t have a clear idea of what they want to convey to begin with.
And all of this is subsumed under the heading of what is valued. Just listen to the way politicians talk, for example. Sloppy, careless, windy, overblown.
In order for people to begin to care about writing, they need to have an audience with sensitive ears.
Absolutely. I have always found that if you don’t really know what you want to say, you stand no chance of saying it, no matter how long you persevere. This really comes out when you do drafting by committee, and it becomes obvious that what you’re really all trying to do is to cover up disagreements on substance using verbal fuzz. So I’ve always told people working for me, and students, that if you know what you are trying to say, the words generally follow quickly. If you don’t, you can go around in circles endlessly. I’d add that the structure of the whole is also important, so that you understand how the sentence you are writing fits into your overall argument. In my case, I try to be 1-2 sentences ahead in my mind of the sentence I’m currently writing. The other thing is that many people have tin ears for how prose–even workaday prose–sounds to others (balance, length, euphony, repetition etc.). I personally never write anything down until I can hear it in my head, and I’m satisfied that it sounds OK. I often go back through what I’ve written making small changes that don’t alter the sense, but do improve the sound. It’s a shame they stopped teaching rhetoric at universities.
“if you know what you are trying to say, the words generally follow quickly.”
Ah yes, old wisdom (Nicolas Boileau, Art poétique, 1674):
Avant donc que d’écrire, apprenez à penser.
Selon que notre idée est plus ou moins obscure,
l’expression la suit, ou moins nette, ou plus pure.
Ce que l’on conçoit bien s’énonce clairement.
Et les mots pour le dire arrivent aisément.
Add in another threat to these, and many other businesses. We have been in a long-cycle of rising, sustained corporate profit margins, and generally expansive fiscal policy in OECD countries. The high profit margins makes it easier to hire firms like McKinsey. As the business cycle rolls over, and governments lean towards austerity, it will make McKinsey seem more like a luxury than a must have.
It can further devolve into a negative spiral if the prospects for adding partners grows dimmer in recruiting and retention as the trade of a really terrible work life for a potential future partnership decrease.
Too bad.
It occurs to me that one of the most important jobs that will be needed going forward will be Fact Checkers for the output of AI. Knowledgeable people who will have the skills and experiences to go through that output and eliminate some of the absolute howlers that AI will make lest the firms – like McKinsey – end up being sued into the ground. And there is the problem. AI is kicking out the career ladders for younger staff so that they have no way to climb. So after the present generation of Fact Checkers retire or go their own way, where is the next generation of them going to come from? The organization of a firm like McKinsey may soon resemble an hourglass. You will have a wide band of underlings at the bottom, AI will be the bottle-neck in the middle, and at the top you will have a large number of upper management and partners trying to skim off as much cash for themselves as possible. I guess that we will see.
Yeah, but the fact checkers can just use ChatGpt… oh wait
I was somewhat involved with external consultants during Thatcher’s reign, as well as with a number of internal government studies. What struck everybody was the inexperience and ignorance of the consultants who were being employed at an astronomical cost. Government departments then had their own internal consultancy units (I worked in several) which dealt with relatively routine issues, headed by some old, wise, individual and staffed by people with perhaps 10-20 years experience in the organisation. Bright young things like me were employed as assistants and note-takers. Large and important studies would often have very senior and experienced teams assembled, quite often headed by recognised experts from outside government. The focus was almost entirely on experience, knowledge and the ability to elicit good ideas and persuade the reluctant. I’d been in the department a good decade before I was entrusted with a study that had any importance.
Outside consultants, even from big-name firms, were often much younger and less experienced–some were recent graduates–and it was never clear to me why we were supposed to take them seriously. The reasons for engaging them were entirely political: they would produce the results and recommendations that Ministers wanted, whether or not they made sense, because that’s what they were paid to do. But most of them had to be shown how to make a cup of coffee, and it was normal for them to brief my colleagues, as the “customer” whilst knowing less about the subject than anyone else in the room. One colleague who was working at the Cabinet Office at the time recalled an exchange with a group of consultants designing a new staff appraisal system (for which there was no proven need), who when asked some fairly fundamental questions about how this or that worked, replied “we haven’t been briefed on that subject yet.” Working with another government some years later, I was asked as an independent witness to verify that, indeed, a firm of computer consultants tasked with writing a report on security threats hadn’t actually answered any of the questions, but just regurgitated their usual clichés, which I was happy to do. When I was asked to be a consultant myself in various parts of the world, it was because I had detailed knowledge and expertise that the government or organisation didn’t have.
All of which makes me think that AI will devastate management consultancy as I have observed it over forty years. Whilst organisations and managements may still feel that they need some kind of outside blessing for what they want to do anyway, or to fight internal battles, what counts is the final product and what it says, not the time and effort that goes into it. So “write me a report that says this merger is a good idea and that it will save lots of money” is a task that can be done quickly by AI, and probably contains no more mistakes, errors and irrelevancies than the kind of report consultants produce. It’s also infinitely cheaper. And I don’t think the spirit is new, either, a long time ago, I remember reading a consultants report about the privatization of some government manufacturing facilities which included a completely bizarre section on how private companies could be profitable but still have inadequate cash reserves. These were the days before word-processing, so I suspect they got two documents mixed up in the photocopier.
We might have encountered some of the same junior consultants. I was in the London office from May to September 1984. I never saw such a bunch of miserable people. They seemed proud of their work on BT and yet I did not hear them mention a single positive outcome.
Your experience with neophyte consultants sounds very familiar. My company paid a rather exorbitant fee to a consulting firm, only to have recent college graduates tasked with much of the heavy lifting. It became quickly apparent that these youngsters did not know how to implement the functionality we needed ahead of time and were learning on the job, leading me to wonder why we couldn’t just do it ourselves.
Or, as my favorite demotivator puts it –
“Consulting: If you’re not a part of the solution, there’s good money to be made in prolonging the problem.”
It would be a damned shame if it turned out “AI” was willing to prolong the problems, but for free.
The version I know is “if it’s not broken, there’s no money to be made in fixing it. So it must be broken.”
Sorry, you have the quote backwards:
“The problem with consulting is you are hired by the problem.”
And “The most profitable clients are the most diseased.”
During my federal civil service career we employed the services a rather well known firm that identifies itselt with its initials, which we came to believe stood for “Keep Paying Me Gladly.”
Sounds like every internal audit team I ever encountered during my tenure in a senior management role. Having a conversation with these — always very young — people (who knew precisely nothing in principle about the work of our organization as a whole, let alone my department) was invariably a pointless waste of everyone’s time. I entertained myself by crossing swords with their executive supervisor at every opportunity during her annual meetings with our senior staff — if my department was having its time wasted and getting nothing out of the exercise, the very least I could extract was entertaining myself by calling her out in front of my colleagues. 😈
I assume their billings were satisfactory …
Queue the South Park Episode with the refrain, “They took our JOBS!!”
Every dollar not paid to salaries and wages inures directly to the bottom line.
There may be millions of unfilled job openings, but I suspect that management is perfectly happy to delay filling them, and keep the present workforce flogged.
With AI, I can’t imagine that there won’t be eye-popping layoffs coming our way. Glad to see both parties and global leaders discussing Universal Basic Income/ Welfare, and planning ahead for the Great Transition.
https://youtu.be/APo2p4-WXsc
“There may be millions of unfilled job openings”
And how many of those are even real, as opposed to part of a phishing operation?
Mckinsey, Bain, BCG et al are about to fall prey to two related foundational tenets of market economies.
1. Unless you’re an entrenched monopoly with pricing power and regulators in your pocket, excess profits in most sectors get competed away (ala “your margin is my opportunity”). The storied histories and prestige of these firms were up to now the bulwarks that kept upstart competitors at bay and the profit mill churning. Alas, the barbarians (AI) are at the gates and have the city surrounded.
2. Technology diffusion has deflationary effects. If tech makes things more efficient, they generally become cheaper over time. An example is “co-pilot” tools reducing the cost of building production ready software by reducing the number of engineers required (while increasing engineering bandwidth).
It is my view that the current all hands on deck response of leaning into the AI trend and deploying bots into the workforce is a forward longitudinal bet attempting to place Mckinsey on the creative side of creative destruction. I also don’t think it’s going to work because it requires taking short-term actions that polish the firm’s tech-forward narrative to a high sheen but are orthogonal to the its longterm survival prospects (eg drastically reducing the amount of juniors coming into the talent pipeline and replacing them with AI bots). It takes competence to evaluate competence, if the pipeline from junior to experienced executive gets cut in half who is going to provide the competent supervision the AI will, in all likelihood, continue requiring when the current crop of senior leaders retires? Without developing these human competencies, the risk of letting confident bots loose on client projects is alarmingly high (a natural consequences of mistaking bot confidence for competence) and said risk comes with high exposure to litigation. Executives inside firms like Mckinsey have no easy choices, they have to navigate quickly to perch themselves on high branches to escape the rising water level below that’s going to drown many a human worker, while at the same time slowly sawing off the same branches by making on-trend decisions that are deleterious to the future survival prospects of their firms.
They will hire experienced players from outside. Which is generally the beginning of the end when it happens, because if you’re not developing your own experts but going to market for them, you truly have nothing to differentiate you and you’re left with nothing to compete on but price. A firm like McKinsey will run on brand fumes for a while, but that won’t last forever.
Count me among those who aren’t terribly fussed at the prospect of their drowning. Couldn’t happen to nicer people, really …
fascinating post and background glimpse of the conventional vs ‘likely’ impending dynamics of C-Suite fracas
If I can take a different tack, are we witnessing the culling of the bullshit job as a consequence of AI? And to take it one step further, is any job which can be performed by an algorithm an unnecessary (and therefore bullshit) job? And to take it still further, are increasingly more and more types of work by this definition becoming bullshit jobs where perhaps they previously weren’t? Will we reach a point, especially in the US where the corporation is legally a person but a tree or an actual person or community has no standing, where the jobs of executives and managing directors can be performed, and performed better, by algorithms? Especially given algorithms don’t have affairs with their HR directors or otherwise embarass the firm?
When I was a coder an accounting department literally, and I mean literally literally, begged me to code an app to do the mindnumbingly boring and repetitive account reconciliation tasks they at the time needed to do. They, and by they I mean the people doing the jobs, not their bosses, argued nobody should be required to do such work, it was inhuman, unnatural, degrading, unethical. So I wielded my coding skills to wipe out an entire section of the accounting department, putting them out of their misery. I know it was merciful, and they had other tasks lined up which they preferred to do, but at the time I felt conflicted and uncomfortable about it. Fast forward to today and I wonder, are the jobs now being replaced by AI of the same category, is it merciful to shift people away from them, away from finance capitalism, toward more meaningful labours, things you can do with your own hands, look at and admire at the end of the day, be proud of, and where you can sleep better?
We’ve gone from armies of corporate secretaries typing on typewriters, to armies of people glued to screens. It has messed with our brains. The digital has destroyed the real. Are we in the process of bringing back the real?
Fact-checking for AI errors strikes me as the ultimate bullshit job. Let me hoist a recent comment from IM Doc to illustrate:
Someone, somewhere, wanted the AI to be more like Dr. House, untainted by human emotions and frailties, wary of human lies and deceit, unconcerned for the consequences, including possibly death if wrong, unconcerned for the costs of tests, only concerned for truth, for “solving for the differential”, and readily willing to flip hypotheses moment by moment if proven wrong but only based purely on the testable facts. Dr House wanted his team to be able to think, to come up with novel diagnoses which he hadn’t thought of, and more importantly which fit the facts. Where they came to a fork in the road where they had no facts to base on, insufficient information, House’s instinct was to do something, some diagnostic test, to generate more observable facts to inform the treatment decision.
But here with IMDoc the AI tool itself seems to have developed a bit of tunnel vision, borrowed human frailties. If I could hazard a guess – the UTI hypothesis may be the least expensive to test for, the AI might have been trained to prioritize tests in order of cost/expense. In which case, yes, it will never be a good doctor. As long as economy is a decision factor it will be run like the economy.
By the way, I don’t know what IMDoc was using but LLM’s remember conversations and build on those previous conversations. If doctor A had a patient with UTI and doctor B comes along and queries the AI on something else entirely, unless there is some way to wipe the memory it’s going to think you’re still on the first patient, will use the details from the first patient and combine with the second, which would be an unfortunate consequence. And may be why the unrelenting focus on the UTI diagnosis.
No, that was not the case for IM Doc’s errors. Many were sheer hallucinations, like names of pharmacies that did not exist (and where there had been no discussion of pharmacies in that session). The notes were also longer than warranted by the note. Far more output than input.
From IM Doc on an AI horror. He sent an image with patient details redacted:
Would the bullshit job here be the recent graduate masquerading as a consultant?
No! You get to wander around and have the client educate you on his nickel.
At least in Europe, a corporation (or an NGO or a professional association or whatever) is a “person” (“personne morale” not “physique”) to give them a legal personality so that they can enter into contracts. What happens if an AI drafts a contract? Who is responsible for errors? What happens if an AI signs the contract on both sides? Is this even possible?
Actually, there’s a bigger danger with all charitable organisations, NGOs, Churches, you name it. Many countries now have extremely complex laws regulating non-profit organisations, and the cost and time involved in drafting and updating the legal documents and having the accounts done professionally can be quite off-putting. Who wouldn’t want AI to do it for them instead? Then what happens if things go wrong?
I think you’re tugging at something important there. And to take your question further, if there are no more human workers because they’ve been replaced by AI then what does a human executive have to offer as an executive? There are no teams to inspire, no internal/interpersonal conflicts to manage, no politics to manage, there is nothing remaining, no intangible, unquantifiable somethingsomething which execs claim as the “forward thinking strategic visionary direction” portion of their jobs, which they imagine themselves to be so good at.
Instead, what might remain is the accountability, the legal entity or construct of the “officer of the corporation”. Someone to sign their selfsame name on the dotted line and be accountable for consequences of wrong decisions.
Except that in this universe corporate executives are generally not accountable, legally or otherwise. Not to the courts, not to shareholders, not to anything or anyone. It is the one abstraction which makes any crime possible and permissible. So there goes the legal personage accountability argument for having human execs and with that gone I can’t think what else they might be good for in an AI dominated workforce.
Correction, I just remembered, execs ARE responsible but only in China or Russia, it seems.
Well, given Luigi’s actions and now the Blackstone CEO getting “unalived”, we might be seeing alot more “accountability” amongst the ruling class. After all it’s very hard to win an appeal against “Madame la Guillotine.”
I can’t remember where I saw the gag but… if you want to have some fun..
the suggestion is:
. join any virtual meeting 30-60 seconds early.
All the AI notetakers of the attendees will be there.
Spend 30 seconds before anyone joins saying how your ship has sunk, you are surrounded by sharks, will someone send help, the life raft has a hole in it and your food supplies are low.
then when everyone else joins behave as normal…
Childish I know but a good way to see if anyone ever reads the transcripts… :D
I think I posted this once already, but there are also issues like this regarding AI Agents.
https://pluralistic.net/2025/08/04/bad-vibe-coding/#maximally-codelike-bugs
The IM Doc comment brought to mind, “Sorry Dave, I cannot do that.” Next week I have my semi-annual 15 minutes with the cardiologist. He is of an age to have been trained pre-computer, but these visits are so programmed and rushed that other than as the go-to source for two prescriptions, I question their utility. I certainly do not want to ever go into a hospital again.
The future abundance of AI advisory resources will also curtail the predatory consulting practice of cultivating embedded relationships, where the consulting firm effectively becomes indispensable to the client through an unending stream of projects concealing dependency on over-priced outside talent. This is a perverse incentive for consulting firms that are run by the numbers. AI advisors will be working within an organization without conflicts of interest.