Another day, another bunch of reasons as to why AI implementations in the workplace fail.
Given that I come across a lot of verbiage about AI, I’ve taken to collecting reasons why sticking an LLM of one flavour or another into the workplace is – with some outlying exceptions – a massive clusterfuck.
Readers may already have read the first dozen or so reasons AI has fallen on its arse in the real world; you know, when people actually start using large language models in practical ways to get stuff done. And heavens-to-Betsy, none of them have anything at all to do with AI itself.
So, hot on the heels of the last batch, here are five more, culled from the marketing output of widget-sellers. (N.B. Without a widget, AI is a bag of cock. So, buy widgets!)
- “Employees aren’t on board, engaged, and ready to work with AI.” If only the thick dullards we select, vet carefully and generally train up weren’t so aggressively against all things AI! They know nothing, despite university educations, much specialist training and real-world experience. They all possess a blind spot when it comes to having a new tool that takes f*cking ages to set up and still f*cks up the simplest of tasks four times out of ten.
Are they insane, or just malcontents who might join a union or ask for paid sick leave? Hang the Luddites! Perhaps they need training on how to use language prompts: - “Trust stalls adoption. A recent survey we paid for states that [make up a low number]% of workers simply don’t trust AI for decision-making.” Why might that be, do we think? Could it be because ‘workers’ have seen the shitshow AI produces, even after intensive training grudgingly provided by ‘decision-makers’ (see above)? Would they trust a new bloke called Colin, recently hired, who is not only incompetent but actively dangerous in terms of permanently f*cking shit up? Maybe they’d trust Colin 5.4, who is physically indistinguishable from Colin 4.2, but takes six times as long as his previous iteration to come to a still-incorrect conclusion, that is, when he doesn’t wander off and stare at a patch of wall for three hours until someone nudges him and tells him to go for a lie down? At which point he reformats everyone’s hard drive.
- “In Asia, AI pilots are running, but there’s a lack of the necessary infrastructure to run AI on.” Yep, those backward Koreans, Malaysians, Japanese, Indians, Chinese and Singaporeans. They’ll never produce phones, cars, electronic devices, software conglomerates, advanced engineering and some of the world’s most educated, multi-lingual, well-trained and capable workers, will they? If only they had better hard drives, fast internet and computer chips they can design and build for us – then, in that dreadful morass of savagery, they might be able to run a large language model. Poor loves. Send them your old ThinkPad.
- “The very pace of model development is too quick for companies to keep up with.” Shouldn’t that read, “the very pace of specious claims made by AI companies is too fast to keep up with”? By adopting every new thing, and trusting in it implicitly, whether or not it exists at all, business success will be yours. Vibe-coding? Get in! Agentic AI controlling your Windows 11 desktop? Why wouldn’t you! Bots talking to bots on social media? How funny is that! This Is The Future, sonny, and artificial general intelligence is just around the corner, when a computer trained on the contents of Instagram and Reddit will possess the IQ of 30,000 PhDs and run on your Motorola. Try to keep up.
- “A lack of execution capacity – the speed at which decision-makers can make decisions.” AI helps you make decisions faster. But can you keep up? Two sugars, or three? Oh fuck, too late. Buy or sell? Come on, come on, faster you tortoise! If only we could choose which pair of sunglasses to buy first, we’d be stylish and almost famous. But dammit, we are too slow to buy stuff as fast as we need it, and we have all ended up ordering swimming pools shaped like a giant cock and balls.
Sadly, this series will be likely to continue. Watch this space!