The Quiet Rot Inside the AI Productivity Story

A wood-decaying fungus recently passed a cognitive test. No brain, no neurons, no consciousness — just a network of threads solving a maze-like problem by sensing chemical gradients and rerouting its growth. Researchers are now asking whether decision-making and memory require anything like what we'd call a mind at all.

It's a genuinely strange finding. But it rhymes with something happening in offices right now, and the rhyme is uncomfortable.

We've spent the last two years pouring money into AI tools on the assumption that intelligence — even the approximate, statistical kind — makes work better. What's showing up instead, in at least some workplaces, is something researchers are starting to call "workslop": the accelerated production of outputs that look finished but aren't really thought through. Emails drafted, reports assembled, slide decks populated — all at speed, all reviewed just lightly enough to ship. The volume goes up. The judgment goes sideways.

The fungus finds the food. The employee hits send. Neither pauses to wonder if the path is right.

The productivity dividend that hasn't arrived

Here's the part that should give pause. Companies are spending billions on generative AI tools, and the aggregate productivity numbers haven't moved in any dramatic way. This isn't unique to AI — the economist Robert Solow made the same observation about computers in 1987, when he noted that you could see the computer age everywhere except in the productivity statistics. It took another decade before the gains showed up clearly, and they required wholesale reorganization of business processes, not just the addition of new tools.

The UNCTAD Technology and Innovation Report for 2025 makes a version of this point at the national scale. For AI to deliver on its promise, countries need robust infrastructure, useful data, and genuine skills — not just access to a chat interface. The spread of electricity didn't transform manufacturing until factories were redesigned around it. The internet didn't transform commerce until logistics, payments, and trust were rebuilt from scratch.

What we're doing right now, in many organizations, is bolting AI onto existing workflows and hoping. That's not transformation. That's electrifying a horse-drawn carriage.

The skill erosion problem no one wants to talk about

There's a subtler issue nested inside this. When cognitive tasks get offloaded to tools before the person doing the offloading has fully developed the underlying skill, something gets lost that's hard to name and easy to dismiss.

Medical residents who use diagnostic AI before they've built intuition may come out the other side less capable of reasoning through ambiguous cases. Junior analysts who've never had to construct an argument from scratch, because the first draft always comes pre-assembled, may struggle when the tool fails or the situation is genuinely novel. The workflow becomes the crutch before the walker was even tried.

This isn't technophobia. It's what happens when efficiency and capability get confused for the same thing. A calculator doesn't make you worse at mathematics if you already understand mathematics. It might, if you never learned to.

What the Scientific Revolution actually required

The 16th and 17th centuries didn't produce a scientific revolution because people suddenly had access to better instruments, though they did. The deeper shift was a change in epistemology: abstract reasoning over intuition, quantitative over qualitative, systematic experimentation over received authority. The tools mattered less than the new habits of mind.

That parallel is worth taking seriously. AI tools are instruments. What shapes whether they make us more capable or less is the intellectual culture around them — whether organizations reward genuine reasoning or just outputs, whether individuals use AI to extend their thinking or to avoid it.

The fungus is impressive. It solves real problems without consciousness. But nobody is putting it in charge of anything that requires judgment about what problem to solve in the first place.

The question for the next few years isn't whether AI is powerful. It clearly is. The question is whether the humans working alongside it are building capacity or quietly trading it away for the comfort of a faster first draft. The answer probably varies enormously by organization, by manager, by person — which means it's still a choice, not a fate.