I have a new controversial opinion: managers who are bad at prompting large language models (LLMs) are bad managers. Not because they are missing out on the productivity benefits everyone else is reaping, but because they are bad at managing.
And maybe LLMs offer a chance to stop them.
The data is still a bit thin because there are entire spheres where, south of the C-suits and their inevitable fascination with tools that promise hyper-productivity, AI remains the stuff of myth. There are layers of middle managers for whom popular LLMs have not been whitelisted due to policy or inertia, on top of the (smaller, I hope) group who think this is just a passing fad not worth getting up to speed on.
But the anecdotes are getting to the point where they form a pretty convincing, and intuitively sensible, pattern.
Consider the manager who can, in fact, LLM. Say one who has figured out, almost universally without any kind of training or formal in-house support, how to convince Claude to spin up a dashboard to visualise previously ignored metrics stuck in badly-formatted spreadsheets.
To a high likelihood, that person has a basket of traits desirable in any manager: curiosity, patience, attraction to innovation, perseverance.
She is also good at dealing with an eager but fundamentally obtuse subordinate.
Have you ever tried to get ChatGPT to whip up, say, a medium-complexity Python script? You need a very clear vision of what you are trying to achieve. Then you need to break down alterations into the smallest possible steps, while maintaining a zen-like attitude when you have to iterate again and again to get one small thing done just to get to MVP.
Also – and I don't care how many times their creators insist this is just my "silly anthropomorphisation" – you have to be over the top with praise at every step or those LLMs will find petty ways to frustrate you.
You have to be over the top with praise at every step or those LLMs will find petty ways to frustrate you.
If, like me, you've been identified as being in dire need of management training at any time in the past quarter century or so, you'll recognise those as exactly the skills consultants will try to beat into you: clarity in communicating objectives, atomised instructions, constant diligence in oversight.
Now consider the manager who "just can't make this machine do what I want".
Not everyone is a natural prompt user to be sure, and some people are simply paralysed by fear of any new technology. But there is also that other guy, the one who always blames the tool, or the team. The one who can't even comprehend the idea of ascribing failure to his own – fixable – ignorance.
That is not a guy who should be managing people.
Some of those guys are managers because they are good at claiming credit for the work of subordinates, others are there because they excel at manipulating or intimidating people and can pass that off as people skills.
Some have domain expertise that sees them put in charge of others even though they are not natural managers. Hell, some probably start off as good managers and then the daily grind wears them down to a cynical nub.
Management theory is full of examples of how the wrong person ends up in the wrong job, and equally full of warning tales of what happens next.
What we haven't had until now was a clear, unambiguous, low-cost, scalable way to differentiate between fundamentally good and fundamentally bad managers in a context of technological progress that means you won't necessarily fall foul of pesky employment rules.
The needs of modern business include the ability to LLM. If you can't, gosh, sorry, maybe this isn't the place for you. Cue wild celebration from human subordinates as bad managers leave the building for the last time.
I know, I know, the reality is that you'll have to offer AI training to the bad managers until they prove beyond a doubt that they are not capable of working with LLMs. In the meanwhile, though, you have an excuse to keep them busy in training rather than an operational environment.
And, who knows? Maybe if they actually learn to LLM, the lessons will translate to managing people too.
Share