Monday Morning Moan - AI 'workslop' is maddening enough for colleagues, but it could be existential for the organization if we're not careful! - diginomica

In Disney’s The Lion King, the evil uncle, Scar, bemoans the fact that he’s “surrounded by idiots”.

I know how he feels.

But – if he were here right now – I’d tell him he doesn’t have it so bad. That dealing with snickering hyenas is less depressing than dealing with lobotomized enterprise employees who have been told to use AI at all costs.

And actually done it.

When my colleague Chris Middleton recently wrote about meat proxies – people who simply relay AI-generated information without really understanding or processing it themselves – it rang a pretty loud bell. Because I was actually experiencing that phenomenon in real time in one of my outside engagements.

And it was genuinely bewildering – and more than a little annoying.

Answers to simple technical questions came back filled with reams and reams of clearly AI-generated text – padded out beyond comprehension with a nauseating mix of ponderous metaphors, faux chumminess, unnaturally decorative turns of phrase and misjudged rhetorical flourishes.

All ultimately wrapped around a point that was... a bit crap – and which could have been said in a couple of lines if the person responding had had even the most basic understanding of the topic.

Each email felt like the emaciated corpse of an idea, wrapped in verbal vomit, inside a rhetorical turd, as Churchill would (probably) have said if he worked in HR.

But the more critical point – beyond the stupidity of it – is that it was just putting the load back onto me – by expecting me to extract my own answer from a flood of text rather than taking the time to think about how to give it to me.

But it came out this week that I am not alone in suffering from this behavior.

Tobias Lütke – an enthusiastic promoter of AI for everything in the early days – recently called this phenomenon ‘slop grenades’ and said he was somewhat appalled by the tendency of his staff to do it. That it was making more work for everyone else.

Other enterprise leaders report the same – with some estimates suggesting that the resulting rework is costing a 10,000-person organization more than nine million dollars a year.

Plus co-pilot licenses, obviously.

I mean maybe when you tell staff they must demonstrably use AI or be fired then perhaps they… do it? Often inappropriately. But then again perhaps that’s just me not really being clever enough to understand the nuances of real-world ‘business strategy’.

But in reality, I don’t think that’s sufficient to explain what’s going on. I think the deeper problem starts with a category error in the way we judge expertise.

Confabulating credibility

As an illustration, there was a ‘wordcels versus shape rotators’ meme a couple of years ago that many people took as an opportunity to ridicule the skills of people who could write – journalists, authors, commentators etc.

Because apparently all those people had become worthless now that AI could produce all the words.

But it turns out people who can write aren't just people who produce words. They are people who can think. Writing is an expression of clarity of thought – and to have clarity of thought you need to have a world model, a structure, a way of seeing things that grows out of personal experience and expertise.

And I think that category error – conflating visible output with the capability necessary to produce it – is at the heart of the enterprise slop we’re now seeing everywhere. Not just in writing – but in every area of the organization from marketing to finance to software development.

Because AI can pretty much produce anything immediately. Things that look plausible, fluent, and convincing – at least if you don’t know anything about the topic. Things which make people more confident in producing and sharing content they don’t understand.

But the more interesting flip-side of this phenomenon – at least to me – is that this also works the other way. Effectively, the plausibility of AI outputs has corrupted the surface signals we previously used to judge competence when evaluating things beyond our core expertise. Fluency, structure and confidence – or lack thereof – used to tell us something useful about the credibility of a piece of work – but AI can now effortlessly produce terrible outputs that nevertheless demonstrate all three.

Which effectively means that execution has become detached from capability – by enabling non-experts to create poor outputs whose fluency nevertheless makes them look credible to both themselves and other non-expert readers.

And so it goes on – compounding and spiraling – until it hits someone who understands the thing – and who then has to de-construct it, work out which parts are usable, and re-do the whole piece of work, which, of course, actually takes longer than doing it in the first place.

So having recently coined Thomas's Law, I'm kind of on a roll. So here’s Thomas's Paradox:

The easier AI makes it to produce expert-looking work, the more expertise you need to know whether the work is actually expert-level.

And the increasingly yawning gap between the first and the second is where slop comes from.

Cultivating capability

Because the real problem here is that to judge something – both during and after production – you have to master the topic you want to judge. And to master a topic you need to practice it for 10,000 hours (yes, yes simplistically – but good enough for this argument, OK). You need to inhabit the problem, develop the skill, fail, fail and try again.

You have to know how to make the thing to understand the thing, to judge the thing, and that’s not knowledge you gain when the finished thing is simply handed to you.

So the people getting the greatest leverage from AI today are often the people who already did their 10,000 hours without it. They know what they want. They know what good looks like. They know when AI has misunderstood them. They know when a plausible answer is missing the point entirely.

And so while AI makes those people extraordinarily productive – because it accelerates execution without replacing their judgment – its plausibility also makes people without that judgement increasingly confident to bypass them.

Which is why, effectively, I believe the enterprise is filling up with slop.

But slop also feels like a signal of where this ends if the organization becomes ever-more dependent on AI and if we don’t give people the opportunities to do the hard yards, the 10,000 hours.

Because today’s judgement in the enterprise is living off the accumulated intellectual biomass of the pre-AI economy – people who acquired their expertise under the old system and were shaped by friction.

Which is why AI makes people with expertise more valuable – even as it makes people without it more dangerous.

But as those people disappear from organizations over time – whether through idiotic schemes to replace them with AI or simple staff turnover – and are replaced by people who never get the chance to do the 10,000 hours, at some point that biomass stops replenishing itself, along with the organization’s ability to apply good judgement.

And that, surely, will ultimately be a much bigger problem for us all than 'workslop', no?