In line with your “take jokes seriously “ principle, how about “the managing director is the only person in the firm who can take a four hour lunch break without damaging production “?
this reminds me of my definition of the class system - "come in to work late and tell your boss that you've got a hangover. if he laughs, you're middle class".
My views on the uselessness/negative vlaue of much management were cemented by managerial enthusiasm for open plan offices (not for them, of course) and resistance to WFH. They love their Panoptican, and don't want any pesky evidence to suggest that the costs they impose on workers are reflected in staff turnover or, to avoid that, wage premiums
I am sort of a middle manager at a law firm, and I react to everything on this list with horror. But the well I personally drink from has been poisoned by various widely reported stories of lawyers delegating tasks to AI and being professionally disciplined and publicly humiliated for it.
Shared on LinkedIn for reach yesterday, handful of comments with answers across the board, from "absolutely none" to someone who used an LLM to summarize their past LLM usage against the 5 questions as criteria to come up with an aggregate score on how likely they would be to use an LLM because "I don't really trust myself." Indeed.
1 & 2 are already happening at scale as you suggest & imho will keep scaling
3 not yet but I can see AI agent to agent comms doing this (initially within strict parameters) pretty soon
4 & 5 not unless you want the machine to be the CEO 😉 … I’ve known senior bankers agonise for an entire morning over the distribution list for an email
And / Or the content of one’s email might automatically initiate an AI HR agent to declare that the sender should be made redundant due to not understanding the core mission and functions of the organization.
The 🤖 HR agent would be also be tasked with performing this function based on a cumulative scoring function for each sender I.e. how many times they send irrelevant and unnecessary messages / queries.
At which point most current managers would cease to exist.
Bureaucratic Enlightenment is coming to understand that email was void all along, that falling over and over again into the cycle of replying is the root of suffering.
I wouldn't be comfortable trusting AI to do any of these things. It could probably do them; but unless the task required immediacy and didn't require a good level of quality and accuracy, why should I? I think that your question is less about the concept of management and more about tolerance of risk
> Rather than having a defined organisation chart, roles and divisions, would you be interested in a model where AI responded to strategic priorities by assigning people to functional and problem-solving teams on the basis of its assessment of their abilities, compatibility and the urgency and importance of the tasks?
Rather than “assigning to,” imagine the AI having a conversation with each employee about available projects and what they would like to work on next. That is, it’s likely that the knowledge needed to do this task well *doesn’t exist” and it needs to be discovered through conversation. In theory, an AI could engage in far more conversations at once.
If the results of all these conversations were converted in proposals to management, would they be any good?
No doubt 4. is the real killer app, so to speak. Based on personal observation, around 80% of middle managers' time is wasted because they each chase 4-5 individual performance objectives of which 2-3 aren't aligned with those of their internal stakeholders and are thus unachievable in a matrix organisation (although this is artfully airbrushed at the time of the performance review). My 80% waste estimate is because a lot more time is spent on doomed objectives, whereas the 1-2 realistic ones tend to be fulfilled by a tide that lifts all boats (eg, market growth). AI could be far more ruthless in assigning realistic objectives to fewer, but more empowered managers.
On current trends, in five years you will not get a choice about any of 1 to 5. Legal will require them all. That's if you still have a job.
Yes, *now* AI is another guy in Chennai. What was it five years ago? What will it be in another five years?
(Don't be misled by the negative press about GPT 5. Many people were upset that it's not as much of an arse-kisser as 4. Those people's opinions are worthless.)
From a business user perspective isn't 4 just a faster Jira ticket creation and assignment flow, and 5 the existing throwing a question into the #analytics Slack channel?
Boy, I am a middle manager, and I can’t see saying yes to any of those other than maybe first drafts of marketing plans, so long as they were carefully reviewed by someone competent before going live to screen out embarrassing hallucinations.
Mmmmaybe hiring hourly workers? While I do do some hiring, the only people I hire who are nominally paid hourly are admins who I and the lawyers I supervise rely on, so screening for intangibles indicating competence and collegiality is big. But for someone who was hiring a larger workforce, possibly?
Procurement, I would assume it would not only be a risk factor for spontaneous expensive errors, it would also be a target for con artists figuring out what its weaknesses were. Again, with a competent human checking everything before any money went out the door, sure. But not without that.
And the last two would be oh hell no, unless maybe I had a Machiavellian plan to make my organization stop relying on email for communication completely.
Question 4, especially, is a great one. I suspect an element of randomness in choosing people would also be welcome (rather than, I assume, just strict taxonomy of skills and past projects).
The "email assigner" of question 5 would be intriguing. Is that really the answer to the problem, though? Don't Slack (and other options) offer a different solution in terms of having channels that anyone can follow?
In line with your “take jokes seriously “ principle, how about “the managing director is the only person in the firm who can take a four hour lunch break without damaging production “?
this reminds me of my definition of the class system - "come in to work late and tell your boss that you've got a hangover. if he laughs, you're middle class".
Or the owner's son.
My views on the uselessness/negative vlaue of much management were cemented by managerial enthusiasm for open plan offices (not for them, of course) and resistance to WFH. They love their Panoptican, and don't want any pesky evidence to suggest that the costs they impose on workers are reflected in staff turnover or, to avoid that, wage premiums
I am sort of a middle manager at a law firm, and I react to everything on this list with horror. But the well I personally drink from has been poisoned by various widely reported stories of lawyers delegating tasks to AI and being professionally disciplined and publicly humiliated for it.
Shared on LinkedIn for reach yesterday, handful of comments with answers across the board, from "absolutely none" to someone who used an LLM to summarize their past LLM usage against the 5 questions as criteria to come up with an aggregate score on how likely they would be to use an LLM because "I don't really trust myself." Indeed.
that's hilarious, thanks so much!
Wrt to professional services space (my world):
1 & 2 are already happening at scale as you suggest & imho will keep scaling
3 not yet but I can see AI agent to agent comms doing this (initially within strict parameters) pretty soon
4 & 5 not unless you want the machine to be the CEO 😉 … I’ve known senior bankers agonise for an entire morning over the distribution list for an email
oh yeah ... I came up with 5 as I was thinking "what's the one thing that everyone would absolutely hate to delegate?"
Receiving the paycheck.
On 5) - Imagine the psychic damage done to users of this tool when the AI decides to send their email to no-one but the void.
And / Or the content of one’s email might automatically initiate an AI HR agent to declare that the sender should be made redundant due to not understanding the core mission and functions of the organization.
The 🤖 HR agent would be also be tasked with performing this function based on a cumulative scoring function for each sender I.e. how many times they send irrelevant and unnecessary messages / queries.
At which point most current managers would cease to exist.
Bureaucratic Enlightenment is coming to understand that email was void all along, that falling over and over again into the cycle of replying is the root of suffering.
.... until the lawyers and investigators show up trying to blame you for something...
I’m going to reference some of your previous posts on AI by taking the liberty of rephrasing your question to the following:
“Would you be willing to outsource this task to a team of dudes in India?”
To which based on your own conclusions in The Unaccountability Machine on outsourcing and black boxes, the answers would be:
1. Possibly but not without giving it a thorough check-through
2 and 3. You might do this but you probably shouldn’t
4 and 5. Not on your bleedin’ nelly mate
I wouldn't be comfortable trusting AI to do any of these things. It could probably do them; but unless the task required immediacy and didn't require a good level of quality and accuracy, why should I? I think that your question is less about the concept of management and more about tolerance of risk
> Rather than having a defined organisation chart, roles and divisions, would you be interested in a model where AI responded to strategic priorities by assigning people to functional and problem-solving teams on the basis of its assessment of their abilities, compatibility and the urgency and importance of the tasks?
Rather than “assigning to,” imagine the AI having a conversation with each employee about available projects and what they would like to work on next. That is, it’s likely that the knowledge needed to do this task well *doesn’t exist” and it needs to be discovered through conversation. In theory, an AI could engage in far more conversations at once.
If the results of all these conversations were converted in proposals to management, would they be any good?
No doubt 4. is the real killer app, so to speak. Based on personal observation, around 80% of middle managers' time is wasted because they each chase 4-5 individual performance objectives of which 2-3 aren't aligned with those of their internal stakeholders and are thus unachievable in a matrix organisation (although this is artfully airbrushed at the time of the performance review). My 80% waste estimate is because a lot more time is spent on doomed objectives, whereas the 1-2 realistic ones tend to be fulfilled by a tide that lifts all boats (eg, market growth). AI could be far more ruthless in assigning realistic objectives to fewer, but more empowered managers.
On current trends, in five years you will not get a choice about any of 1 to 5. Legal will require them all. That's if you still have a job.
Yes, *now* AI is another guy in Chennai. What was it five years ago? What will it be in another five years?
(Don't be misled by the negative press about GPT 5. Many people were upset that it's not as much of an arse-kisser as 4. Those people's opinions are worthless.)
From a business user perspective isn't 4 just a faster Jira ticket creation and assignment flow, and 5 the existing throwing a question into the #analytics Slack channel?
Boy, I am a middle manager, and I can’t see saying yes to any of those other than maybe first drafts of marketing plans, so long as they were carefully reviewed by someone competent before going live to screen out embarrassing hallucinations.
Mmmmaybe hiring hourly workers? While I do do some hiring, the only people I hire who are nominally paid hourly are admins who I and the lawyers I supervise rely on, so screening for intangibles indicating competence and collegiality is big. But for someone who was hiring a larger workforce, possibly?
Procurement, I would assume it would not only be a risk factor for spontaneous expensive errors, it would also be a target for con artists figuring out what its weaknesses were. Again, with a competent human checking everything before any money went out the door, sure. But not without that.
And the last two would be oh hell no, unless maybe I had a Machiavellian plan to make my organization stop relying on email for communication completely.
Question 3 is kind of (albeit the other way around) the basis of USA vs Realpage https://www.justice.gov/archives/opa/pr/justice-department-sues-realpage-algorithmic-pricing-scheme-harms-millions-american-renters
Question 4, especially, is a great one. I suspect an element of randomness in choosing people would also be welcome (rather than, I assume, just strict taxonomy of skills and past projects).
The "email assigner" of question 5 would be intriguing. Is that really the answer to the problem, though? Don't Slack (and other options) offer a different solution in terms of having channels that anyone can follow?