Product design needs to stop producing information and start designing decisions

The problem has a name: workslop

A 2025 study by BetterUp Labs with the Stanford Social Media Lab found that 40% of US office workers had received workslop in the previous month: AI-generated content that looks finished but doesn't solve anything. On average, each episode consumed nearly two hours from whoever had to clean up the mess.

Do the math for a ten-person team, one piece of workslop per person per month: that's about 20 hours, more than two business days, just cleaning up someone else's shortcut. Designer Patrizia Bertini uses this data, in an essay published on November 13 at UX Collective, to open a question that matters directly to those who design products: if AI produces information faster than any team can read it, the bottleneck is no longer generating content. It's deciding what to do with it.

This is the shift I think our field still hasn't fully processed. Discovery (the research that informs what to build) has always been about producing knowledge from scarce evidence. When evidence becomes abundant and cheap, the work of those who design products stops being about generating information and becomes about designing the process that turns that volume into a reliable decision.

More data isn't more wisdom

Bertini draws on an old model from information management, the DIKW pyramid (Data, Information, Knowledge, Wisdom), formalized by Russell Ackoff in the late 1980s: data is the raw signal, information is organized data, knowledge is information interpreted and applied, wisdom is what allows judgment between options. Each level depends on the one below it, but the sum of knowledge doesn't turn into wisdom on its own.

In 1987, Milan Zeleny proposed the same ladder in business language: know-nothing (data), know-what (information), know-how (knowledge) and know-why (wisdom). It's this last level that allows comparing and judging decisions, not just executing them.

Wisdom is always knowledge, but knowledge is not always wisdom.

Jacques Steyn, quoted by Patrizia Bertini in an essay at UX Collective

The practical point for those who design products: AI delivers know-what and know-how at a speed no research team ever had before. Know-why, which is exactly what guides roadmap priority, remains human work, and it's the work that the rush to use AI tends to cut first.

What AI solves, and what it pushes onto you

Even know-how doesn't come ready-made. Vectara's hallucination leaderboard, which tests how much models make up when summarizing a document, registered in 2026 rates between 1.8% and 24.2% depending on the model, according to Bertini's article. These are numbers specific to one task (summarization), not a universal error rate, but they show that even the lowest level of the pyramid requires checking.

This hands back to product designers a responsibility that the AI speed narrative tends to hide: deciding where human verification is mandatory and where it's wasted effort. It's a product decision, not an implementation detail, and usually nobody on the team is deliberately designing that flow.

Decision theatre or decision design

The phrase that titles Bertini's essay describes a trap I recognize in a lot of AI roadmaps: putting an overloaded person in front of an AI output and calling it "human in the loop" doesn't guarantee any decision at all.

Put a disengaged, overloaded or alienated person in front of an AI output and you get decision theatre: the illusion that a decision happened when it did not. Someone clicked "approve". Nobody decided.

Patrizia Bertini, designer and author of the essay at UX Collective

To become a real decision, not theatre, Bertini lists three conditions that I read as a product brief: competence to read context and separate a valid answer from a well-formatted one; authority (name, title, the power to override AI and answer for the outcome); and a process designed to give the right time and information, without drowning the decision-maker in decision fatigue. Miss one of the three, and the decision collapses into theatre.

The counterargument worth taking seriously

One can disagree here, and the objection is strong: not every decision needs expensive human oversight, and insisting on it for low-risk decisions is decision fatigue disguised as responsibility. Article 26(2) of the European AI Act itself, cited by Bertini, doesn't require human oversight for every use of AI, only for systems classified as high-risk, such as loan approval, school placement, or access to benefits.

This undercuts the reading that decision design means putting people in to review everything. It means the opposite: deliberately designing WHERE human review is mandatory (because it affects someone's rights and the company can't hide behind the system) and where it only gets in the way. That distinction is also product work, and usually no PRD treats it as a design decision.

What changes for those who design products in Brazil

If yesterday's discovery was about producing knowledge from scarce evidence, today's discovery is about designing the flow that decides what to do with abundant evidence. Bertini points to three families of skill that become the work itself, not an extra: critical judgment (being skeptical, asking where a claim comes from, noticing what's missing), synthesis (organizing knowledge into a structure that increases reliability), and human skills (empathy, the courage to disagree with the machine, putting your name behind the outcome).

The good question to bring to your own roadmap isn't "where do we use AI". It's: in the decisions your product makes today, who has the name, title, and authority to answer for them when AI gets it wrong, and does the process around that person give real time to decide, or just an approve button?