How to Spot and Stop
Discovery Slop

​Dear Reader,​

The first time I heard Julia mention the idea of Discovery slop, I knew she was onto something.
​
Discovery Slop is research that looks shiny but is sh*tty. AI-generated insights that sound compelling but are built on summarized contradictions.
​
Prototypes that test usability when the actual question was strategic viability.
​
Output that's impressive to stakeholders but doesn't reduce any meaningful uncertainty.

I keep seeing it in my own work with product teams. The output volume has gone up. The quality of decisions hasn't followed. Engineering builds faster, but on thinner foundations.

That's why I wanted to share the key takeaways from last week's webinar she and I delivered:

#1 DO's and DONT's of AI Research Agents

DO:

• Access raw data quickly ("Can you give me an overview of the biggest pain points also give me quantitative insights about how many people mentioned it?")
• Information about one customer ("What has this customer "xyz" mentioned about errors that occur in the calculation?")

DON'T:

• Copy & paste the content
• Use the content without going into the raw data
• Just use the agent without further interviews

#2 AI Tightens the Trio

Everyone in product loves a Venn diagram. So, naturally, I used one to describe the change in collaboration between the usual trio: PMs, designers, and engineers keep their titles, but the overlap between them grows because the cost of stepping into another discipline has collapsed.

But how does that expansion look like, and what remains?

A PM can ship a low-key bug fix through Cursor in seven minutes (Julia did exactly that the day before our session). An engineer can extract customer call patterns from a research agent without scheduling a discovery cycle.

Which raises a different question for each individual: not "what new skills should I learn?" but "where is my unfair advantage, and what am I doing to make it deeper?" For product, this is about providing direction. For engineers, it might be ensuring functionality. Appearance and system for designers. The roles should specialize more next to the expansion into commodity skills through AI.

#3 Synthesizing is not summarising

Julia made a distinction during the webinar that I want to lift out, because it does most of the work:

Summarising compresses what is already in front of you. Synthesizing connects what you have absorbed from a dozen different contexts that were never in the same room. The first is a function. The second is a person who has been paying attention.

AI is excellent at the first. It has no access to the second.

#4 The Skill that Remains

Human Judgement. Period.

An over-reliance on AI will lead to an erosion of your mental models and gut feelings, which will remain critical (especially during accelerated building speeds).

If you want to dive deeper into this topic, I recommend following Julia on LinkedIn and considering her in-person workshop on AI Fundamentals for Product Managers.

Thank you for Practicing Product,

​Tim​

If you consume one thing this week, make it this...

Who is Tim Herbig?

As a Product Management Coach, I guide Product Teams to measure the real progress of their evidence-informed decisions.

I focus on better practices to connect the dots of Product Strategy, Product OKRs, and Product Discovery.