So Far We Would Only Expect Islands, Not Continents, & Even Those Knee-Deep Swamps of "WorkSlop"
A Wednesday Morning Take on Why We Would Have Expected to See Business-AI Enthusiasts Disappointed at This Stage: The dynamo, the computer, & now today’s GPT (this use; General-Purpose Transformers) LLM MAMLMs: “workslop” is not new, GPTs (this use: General-Purpose Technologies) pay out only after complementary investments & organizational change, for the gains come with redesigned work flows—governance, roles, and standards—not those that still use old pipes, plus add platform moats maintained by angry Leviathans: I see seven reasons not to have expected things to work out, so far, much differently than they have…
Gary Marcus makes a catch:
Gary Marcus: Why is the ROI on Generative AI so poor? <https://garymarcus.substack.com/p/why-is-the-roi-on-generative-ai-so>: ‘Excerpt from a new study from BetterUp labs and researchers at Stanford: “So much activity, so much enthusiasm, so little return. Why? In collaboration with Stanford Social Media Lab, our research team at BetterUp Labs has identified one possible reason: Employees are using AI tools to create low-effort, passable looking work that ends up creating more work for their coworkers. On social media, which is increasingly clogged with low-quality AI-generated posts, this content is often referred to as “AI slop.” In the context of work, we refer to this phenomenon as “workslop.” We define workslop as Al generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task…”… [Plus:] ““Of 1,150 U.S.-based full-time employees across industries, 40% report having received workslop in the last month”. Just think how high that percentage can go!…
What do I think of this? I think it confuses things but tangling seven different threads into a single ball of yarn. They are different things, and need to be distinguished:
(1) White‑collar work is acquiring, filtering, processing, and outputting information. When tools boost output volume without raising signal, they tax the system: more to scan, triage, reconcile, and discard. We have seen this: email’s firehose, the web’s sources, Slack’s atomized threads, dashboards about dashboards. Generative AI adds drafts, summaries, and “insights” that look polished but are thin, duplicative, or off‑task. A product team gets five “AI‑assisted” market summaries weekly, each wrong about competitor pricing. A finance group receives daily “insights” that restate filings without analysis. Attention does not scale with content; cognitive load rises nonlinearly as options multiply and coordination costs mount. Without gatekeeping, standards, and workflow redesign—schemas, editorial layers, curated repositories, incentives for fewer but better outputs—the marginal document lowers productivity by adding acquisition and filtering work. Think plumbing, not magic: a poorly tuned pump increases flow and turbulence, reducing pressure at the point of use. You would not expect a one-for-one benefit in terms of the output of the system, not at all.