Generating text saves time in the moment — and wastes it in the long run
This post was hard to write
It hurt my brain, strained my eyes and put a crick in my neck. On occasion, I’d give up, go to the window and watch a storm blow through the neighbourhood (seagulls unfazed; pigeons bedraggled). Puddles would swell and a sentence would float to the surface of my mind like a bright bubble, forcing me back to the keyboard so I could pop it on the screen.
That’s how I form a complex argument. Line by line. In fits and starts. It’s the only method I know. And I’m in good company. Greater minds than mine have used writing to explore worlds both inner and outer. Investor Warren Buffett claimed there was nothing like it to ‘get your thoughts straight’. Journalist Joan Didion wrote to find out what she was looking at and ‘what it means’. Essayist Michel de Montaigne believed writing could save him from ‘going mad from the contradictions’ he found among mankind.
Today, I’m getting my thoughts straight about AI. I’m looking at businesses losing their minds over workslop and wondering whether writing can restore judgement. Let’s examine a few of mankind’s contradictions, shall we?
Slop became ‘workslop’
So… slop. How do we feel? I’m embarrassed to admit I snorted when I first saw shrimp Jesus. The concept of Fruit Love Island put a grin on my face. But my amusement at AI-generated guff ceased when it reached the workplace. Half of managers have received ‘workslop’ (Stanford Social Media Lab and Betterup, 2025), usually in the form of artificially-generated documents and messages. Paragraphs of text that suck time, without contributing anything meaningful to discussion inside the organisation.
‘Someone sent me what was clearly an AI generated doc. I read it, realised it was all AI, had my AI summarise it and ended up with nothing that brought new thinking to the idea… This is a doom loop. If we’re not careful, we’ll end up with a very expensive way to avoid writing and reading in an ocean of sameness, genericness…’
– Ryan Roslansky, executive vice president at Microsoft
Tasks would go quicker – on average 20 minutes quicker – if the sender did the work themselves (Stanford Social Media Lab and Betterup, 2025). Sure, the sender rips through their to-do list, but then others have to put in extra effort to read, check, challenge and correct the mucky stuff.
Workslop became ‘slop grenades’
Having seen the devastation first hand, Shopify CEO Tobias Lütke calls artificially-generated work ‘slop grenades’. They land without warning and blow your day to hell. Artificially-generated content slows down projects, reduces team efficiency, and makes work less meaningful, according to half of employees (Stanford Social Media Lab and Betterup, 2025).
It’s significant that Lütke is leading the campaign against slop. As the top cat at Shopify, he made reflexive AI usage ‘a baseline expectation’ for employees and built an index of one billion products so AI agents could seek retail therapy. He’s said things like, ‘AI will totally change Shopify, our work, and the rest of our lives.’ So he’s unlikely to about-turn on adoption. However…
Other businesses have slowed their rollouts
Meta scaled back plans to replace staff with AI agents. Microsoft and McKinsey set limits on the use of AI tools. Still others invented processes to keep a ‘human in the loop’. At ING, product teams must document how AI tools preserve human judgement. JP Morgan asks employees to stretch their brains by drafting their own hypotheses before going to the bank’s AI tools. Teams at P&G come together from various functions to check AI’s handiwork. Fact-checking is routine at Zapier.
‘I don’t need to know that gen AI helped you write a PowerPoint deck, but I do need to know that you’ve read and fact-checked every word in that deck. The same goes for me: When I use gen AI, I decide whether to use what it produces, and I must stand by everything in the message.’
– Wade Foster, co-founder and CEO at Zapier
These organisations aim for something greater than efficient activity. They hope to produce advanced thought. As we’ve seen, that means putting ‘humans in the loop’, a fancy phrase for asking employees to think critically and apply judgement, certainly through reading AI output before passing it on, often through (gasp) writing themselves.
Writing has a 450-year track record of improving thought
In the 16th century, Michel de Montaigne became the first writer to call his work ‘essays’. He chose the label because, in French, ‘essayer’ meant an attempt, test or experiment. That was all he meant to do. Each essay was an attempt to understand, a test for his powers of reasoning and an experiment as he flung himself at a topic, unsure of what he’d find.
450 years later, Jeff Bezos pushed employees to take up long-form writing for similar purposes (‘better thought and better understanding of what’s more important than what, and how things are related’). Amazon managers learnt to write six-page memos – a change described as the company’s ‘secret sauce’. Competitors relied on bullet points, which were clearly inferior.
‘Full sentences are harder to write. They have verbs. The paragraphs have topic sentences. There is no way to write a six-page, narratively structured memo and not have clear thinking.’
– Jeff Bezos, former president and CEO at Amazon
The world has changed since Montaigne published his essays. It’s changed even since Bezos dismissed the conventions of PowerPoint slides. AI has developed to the point where you can generate a six-page, narratively structured memo without thinking. But doing that would destroy the value of the exercise. You’d be gift-wrapping a grenade, wasting your colleagues time and the chance to do some real thinking.
Isn’t that what we’re supposed to be doing anyway?
We used to go by ‘knowledge workers’. Knowing was as much of the job as prompting. And with enough collective will to change the workplace, it could be again. Two-thirds of employees are nostalgic the way things were before AI (Adaptivist, 2026). They just need arguments to deploy against Shopify-style mandates.
Here’s one: the benefits of reading, understanding and using information go beyond the task at hand. If the information sinks into your mind, you get to carry that memory forwards, into phone calls, meetings and dinner parties. But an intense working relationship with AI stops these memories from forming.
MIT scientists asked students to write essays while wearing funky brain scanners. The students were divided into groups. Some wrote without help; others got ChatGPT. There was a big impact on brain activity. Students using ChatGPT were barely processing the information. When students were asked to quote their essays, GPT-users couldn’t, while unaided writers did just fine (Kosmyna et al., 2025).
Here’s another argument: knowledge leads to sound judgement, which continues to be an advantage – even with AI on your phone. That was proven through a study of Kenyan entrepreneurs who were given chatbots that provided a list of ‘tailored, actionable, enumerated strategies’ that the businesspeople could interrogate for more tips.
The effect was striking. AI helped entrepreneurs who were already successful and hurt those who were already underperforming (Koning et al., 2026). Journalist Steven Melendez explains:
‘[T]he difference wasn’t between the types of questions users asked the AI or the responses they received in return, which weren’t significantly different between high and low performers. Instead, it appeared that the high-performing users selected different advice from the AI to follow.
‘In many cases, they followed recommendations specific to their businesses, including one entrepreneur who took advice on buying a generator to deal with rolling blackouts in his area; one who began selling cold sodas to patrons of his car wash; and one who received guidance on what sort of chickens to buy for an agricultural business…
‘By contrast, the low-performing entrepreneurs were more likely to take more generic suggestions, such as lowering prices or focusing on advertising. The trouble is that those more general tips can’t address deeper business issues and can actually backfire, because advertising and discounts cost money that may not be recouped without other strategies to grow a business.’
So at the very least, enter your chat with a plan
The MIT study showed there’s a role for these tools. When students wrestled with ideas on their own before revising with GPT, brain-wide connectivity surged (Kosmyna et al., 2025). So here are my practical recommendations:
- To grow as an expert, keep hold of that first draft.
- To own your work, do the final check as well.
- To find meaning in the process, learn as you go.
‘Writing isn’t just the production of sentences. It’s the training of endurance by way of sustained attention. It’s a way of learning what one thinks by attempting to say it. An LLM can reproduce the appearance of that activity, but it can’t replace it, because the value lies not only in the object produced but in the transformation that occurs during its making.’
– Micah Nathan, lecturer and author of novels Gods of Aberdeen and Losing Graceland
Ah, the weather’s cleared outside my window
The rain’s stopped. I will too. This essay is enough of an attempt at understanding. Drafting this has forced me to articulate and defend my beliefs. I’m happy with this half-formed thing which, as you’ve read this far, now overlaps my mind and yours. Perhaps we can discuss it and I’ll reconstruct this essay with better judgement. There’s nothing final here, just a cycle of discovery that runs counter to the doom loop. I’ll write soon.